The first Digitalization, Data Science and AI conference took place on February 1-2, 2024. Below you can find the conference program.  

February 1, 2024


Welcome by: 

  • Sofie Castella, DDSA
  • Serge Belongie, P1
  • Thomas Riisgaard Hansen, DIREC
  • Toine Bogers, IT University of Copenhagen / P1



AI’s increasing role in human-to-human communication and what it means to all of us

In Mor Naaman’s talk, you are going to hear about some fun experiments he and his research group did to understand AI’s impact on communication. One in which they documented what they call the replicant effect and another one where they showed how interacting with AI autocomplete can impact not only what you write but also your opinion.

Naaman’s talk will give you insights that will help you rethink how you develop and deploy AI applications.

About Mor Naaman

Professor of Information Science at Cornell Tech where he leads the Social Technologies research group and serves as the associate dean for faculty affairs. Mor Naaman’s research focus is on the trustworthiness of our information and communication ecosystem, including the impact of AI.

ROOM: Sal C + Vandrehal mod park

ROOM: Salon 22

Topological Data Analysis provides a framework for extracting and analyzing geometric and topological features of data. This relatively young field has shown promise in a range of applications, including the characterisation of disordered materials, biological cells, plant leaves, bones, cancers, DNA, medical conditions, and spatial data. The ability of TDA to generate a summary of the shape of data interfaces nicely with a range of machine learning and artificial intelligence pipelines, not only providing features to use, but also increasing the efficiency and effectivity of such algorithms by reducing the number of features required.

Due to the mathematical underpinnings that lend TDA its power, understanding which tools are suitable for particular data sets and interpreting the output requires a significant knowledge base, which is difficult to obtain in isolation. This session aims to introduce a wide range of researchers to a variety of available TDA tools, in particular publicly available software, to enable them to decipher which tools are suitable and use the results to gain new insights. By introducing TDA broadly in the first part of the session, attendees will be equipped with the necessary knowledge to test TDA on some data under the guidance of invited domain experts. Participants will be exposed to data sets which are well understood, as well as provided the opportunity to bring along their own data to test and discuss. They will be encouraged to suggest types of data beforehand, allowing the domain experts to tailor the session to the participants’ interests.


  • Introduction to TDA | C. Biscio, AAU | 30 mins.
  • Applying TDA to Data | Interactive time where attendees will use tools from TDA to analyse some data, via provided python jupyter notebooks | guided by C. Biscio, L. Fajstrup, S Sørensen, A. Svane, Y. Bleile | 45 mins.
  • General discussion and questions | 15 mins.

Target Audience

  • Any researcher interested in using TDA, up to 20 participants.


  • Christophe A.N. Biscio, AAU
  • Lisbeth Fajstrup, AAU
  • Søren Strandskov Sørensen, AAU
  • Anne-Marie Svane, AAU
  • Yossi Bleile, AAU


  • Yossi Bleile, AAU
  • Matteo Pegoraro, AAU
  • Anne Marie Svane, AAU

ROOM: Salon 23

This workshop focuses on theory and method in human computer interaction.

Discussion of theory has always been central to HCI. Recently, we have seen an increasingly thorough and extensive collection of data about interactive computing and in many ways new technologies, usage contexts and user groups are constantly emerging. As HCI
researchers, we are increasingly challenged to make interactive systems usable and useful.

These developments suggest a need to return to discussions of theory, the generalizations it allows, its role in design, and the difficulties in creating, using, and testing it.

In this 90-minute session, we approach these questions in two ways. First, we have a serious set of invited short presentations on issues surrounding theories in HCI. Their framing, possibilities, limitations, the ways they are developed and/or used. The presentations will
present different theoretical approaches within HCI and discuss the different roles which theory can have within the field.

Second, we discuss these issues with the entire audience in an engaging and interactive

Target audience

Junior and senior researchers in human computer interaction.


  • Professor Susanne Bødker, Department of Computer Science, Aarhus University
  • Professor Kasper Hornbæk, Department of Computer Science, University of Copenhagen
  • Professor Pernille Bjørn, Department of Computer Science, University of Copenhagen

ROOM: Salon 30

The session provides a platform for PhD students working on Embedded AI to present the recent developments and results of their research, fostering knowledge exchange and collaboration among their peers, as well as engaging with participants from the Danish academic and industry communities interested in this field. The theme covers the development and deployment of machine learning frameworks (embedded inference and/or on-device training) for resource-constrained devices such as industrial pumps or hearing aids.

In addition, this session will also host the DAIS Open Day, whereby the DAIS Danish cluster will share the highlights of the DAIS project. DAIS (Distributed Artificial Intelligence Systems) is a pan-European KDT JU project bringing faster, more secure and energy efficient data processing solutions through the development of edge AI software and hardware components. DAIS has a consortium of 47 partners and is organized in national clusters. The contributions of the Danish
Cluster are centered around a Demonstrator, led by Danfoss Drives, titled “Frequency converters as edge nodes in industrial automation”.

Session format and program

The session is organized as a workshop consisting of a series of presentations, followed by interactive discussions and Q&A sessions. The finalized program including confirmed speakers/presenters is reported in the following table. 

  • Opening introduction | Session organizers | 5 mins. 
  • Business Models for AI of Things (AIoT) | Reza Toorajipour, CBS | 10 mins. 
  • Embedded AI | Amin Hasanpour, DTU | 10 mins. 
  • Computationally-efficient State Estimation of Degradation | Iman Sharifirad, AU | 10 mins. 
  • Compilation and Optimization of AI Applications | Alessandro Cerioli, DTU and GN Audio | 10 mins. 
  • Overcoming Computational Costs in Deep Learning for Audio Processing | Riccardo Miccini, DTU | 10 mins.  
  • Data Cleaning for Sensor Data in Edge Sensor Nodes | Xiao Li, RUC | 10 mins. 
  • Distributed AI Systems (DAIS) Open Day | Christian Uldal Graulund, Danfoss Drives, Felix Blaga, Octavic, Xenofon Fafoutis, DTU | 20 mins. 
  • Closing and wrap-up | Session organizers | 5 mins. 

Target Audience

The target audience is members of the academic and industrial community of Denmark with a general interest in AI-driven embedded, cyber-physical, and IoT systems, and a special interest in Embedded AI and TinyML. The target audience includes all members of the DIREC WS6 on “Cyber-Physical Systems, IoT and Autonomous Systems”. 


  • Luca Pezzarossa, DTU Compute
  • Xenofon Fafoutis, DTU Compute
  • Riccardo Miccini, DTU Compute and GN Audio
  • Christian Uldal Graulund, Danfoss Drives A/S
  • Felix Blaga, Octavic
  • Fabian Fernando Jurado Lasso, DTU Compute
  • Juha Kuusela, Danfoss Drives A/S

ROOM: Salon 25

Artificial intelligence, driven largely by rapid advances in deep learning technology, has produced exciting results across various scientific disciplines and practical applications over the last decade. As generative models become more sophisticated, their role in automating the generation of software will likely grow, e.g., GitHub Copilot, Amazon CodeWhisperer, or OpenAI Codex. However, like all technological advancements, it’s crucial to approach this domain with a balanced perspective, appreciating its vast benefits while remaining vigilant of its inherent risks.

The conference session on “Safe and Trustworthy AI-generated Software” aims to explore the latest advances and challenges in developing generative AI for program synthesis and automatic programming. With the widespread adoption of AI-powered systems across various industries, the need for trustworthy and reliable AI-generated software has become increasingly important. It emphasizes the interconnected fields of formal methods, program synthesis, automatic programming, generative AI, and hybrid approaches.

Topics of discussion include using machine learning algorithms for program synthesis, developing generative models for automatic programming, and ensuring the safety and trustworthiness of AI-generated software. Additionally, we aim to include the ethical considerations surrounding the use of generative AI for software development, including data privacy, bias, and accountability issues.

Topics of interest

  • Program Synthesis and Automatic Programming: Modern AI systems can be designed to automatically derive programs from high-level specifications, eliminate the traditional need for manual coding, and help create highly optimized code.
  • Generative AI: Models and systems that can create content, be it code or otherwise, traditionally generated by humans and the principles guiding these AI systems.

Benefits of Generative AI in Automatic Generation of Programs 

  • Efficiency: Generative AI can produce software at speeds incomparable to manual programming, accelerating development cycles and product launches.
  • Optimization: With the right training, AI-generated software can potentially be more optimized than human-written code, as AI can evaluate billions of permutations to find the most efficient solution.
  • Customization: Generative AI can produce tailor-made software solutions for specific tasks or environments, catering to niche needs without extensive manual intervention.
  • Resource Allocation: Automating the coding process can free up human resources to tackle more pressing, creative, or complex challenges.

Risks and challenges

  • Safety Concerns: AI-generated software may inadvertently introduce vulnerabilities or bugs that are challenging to detect and fix. The unpredictability of some AI models can lead to unstable or unsafe code.
  • Ethical Implications: The ability of AI to generate software raises questions about intellectual property rights, job displacement, and the moral considerations of delegating such critical tasks to machines.
  • Dependence on Training Data: AI models are only as good as the data they’re trained on. Poor quality or biased data can lead to flawed or discriminatory software.
  • Interpretability: AI-generated code might be efficient, but it could also be hard for humans to understand, making debugging or modifications challenging.

Overall, this session will provide a unique opportunity for researchers, developers, and industry professionals to discuss insights into the latest advances and challenges in the development of safe and trustworthy AI-generated software.

Target Audience

The session aims to bring formal methods, software engineering, trustworthy systems, and AI researchers interested in generative AI into one session.

We expect 15-30 participants.

Furthermore, we refer to the related DIREC workshop “Verifiable and Robust AI (VRAI)” in Sandbjerg on November 6-10, 2023. 


  1. Welcome | 5 mins.
  2. Lightning round | 15 mins. 
  3. Invited talk by Professor Marta Kwiatkowska, Oxford University | 15 mins.
  4. Talk by Tenure-Track Assistant Professor Christian Schilling, Aalborg University | 10 mins. 
  5. Talk by Professor Thomas Hildebrandt, University of Copenhagen | 10 mins. 
  6. Talk 3 | 10 mins. 
  7. Discussion round | 25 mins. 


  • Kim Guldstrand Larsen, AAU
  • Boris Düdder, KU
  • Zheng-Hua Tan, AAU

ROOM: Auditorium

As deep learning is becoming a more and more integral part of many companies’ workflows and people’s everyday life, the need for data increases exponentially. Moreover, as more specialized tasks
are automated through AI, the required data is becoming more specific and general open-source datasets do not contain the needed information in large enough quantities. Deep learning tasks such
as object detection, segmentation, depth estimation, anomaly detection, among others greatly benefit from transfer learning and additional tuning on specialized data.

The problem is that a lot of the time gathering specialized data can prove costly and time-consuming, as creating such datasets can take many months and even years and requires the purchase and
installation of costly hardware. In other cases, capturing footage can be deemed outright dangerous or impossible without the help of specialists or in very small intervals of time.

This leaves a large gap in how such data can be captured, with current research focusing on modeling 3D digital twins and using them to capture data, augmenting synthetic elements in real datasets or
using generative networks to synthesize the required information. These research fields perfectly encapsulate the modern need for combining knowledge from deep learning, computer vision,
computer graphics and interactive systems, to create functional and useful systems for synthesizing and augmenting such data.

This parallel session will aim to bring together researchers and PhD students working in the intersection between computer vision, deep learning and computer graphics. By showcasing their
current work through a series of short talks we aim to garner a better understanding of the work done in this field in Denmark and establish the possibility for collaborations and knowledge sharing.
We hope to be able to establish a network of researchers from universities and companies interested in or working in the fields of synthetic data generation and augmentation in Denmark.

Researchers that would like to showcase their work should send their application to Ivan Nikolov ( containing:

  • Name and affiliation
  • Area of research and work
  • Topic of the proposed presentation
  • Up to 1 page abstract describing the topic of the presentation.
  • Deadline is 14.01.2024

Session Format

  • The parallel session will include a 30-minute keynote speech by Dr. Barry Norton, VP of Research at Milestone Systems, explaining the company’s experience with synthetic data
    from a responsible tech perspective. Dennis Schou Jørgensen, Principle Software Engineer at Milestone Systems, will give a demo of these works.
  • The second part of the parallel session will be separated into 10–15-minute presentations from participants in the D3A conference that are interested in sharing their research in the
    field. A call for participation will be sent out through relevant channels. Depending on the number of participants, each person will be given 10 minutes to present their work plus 5
    minutes for questions.
  • The session will end with free time for discussion between guests and participants and the possibility to form new networking opportunities. 

Target Audience and Size

The target audience are researchers and PhDs both from the industry and university that are using or generating synthetic data, as well as people interested in utilizing synthetic data to solve problems
that they are facing like lack of specialized datasets, not enough data or data variations. We envision that the session will be of interest to both deep learning and computer vision specialists, as well as
researchers working with computer graphics, interactive applications and even game development. We will have to post an invitation to researchers in Denmark working with generating and using
synthetic data to present their work. The time constraints dictate that there will be maybe time for 3-4 talks of 10-15 minutes, plus a possible 30-minute invited talk. Other than that, no maximum caps of
participant listeners are needed, as people can come to listen in on the talks.


  • Dr. Barry Norton, VP of Research at Milestone Systems
  • Dennis Schou Jørgensen, Principle Software Engineer at Milestone Systems
  • Chosen researchers that will present their research topic in 10 to 15 minutes as part of the second part of the session.


  • Kamal Nasrollahi, Director of Research, Milestone Systems/ Professor of Computer Vision and Machine Learning, Aalborg University
  • Ivan Nikolov, Assistant Professor of Computer Graphics and Computer Vision, Aalborg University


Today’s automation by means of robots to a large extent relies on manual processes and case by case programming. Therefore, the establishment of automation solutions usually takes long time and requires high level engineering expertise that needs to be built up during extensive university educations as well as practical experience. This makes the process of deploying solutions slow and expensive. Altogether, this severely limits how companies can adopt robotic solutions cost-efficiently.

In this session, we will discuss how to enable robot solutions that can capture the knowledge embedded in already running automation solutions and make it available for fast engineering of new efficient
automation solutions. Our vision is to use Data Science Methods to enable the gathering of robot data of already established solutions in a data base and modeling robot processes in simulation. Given a new robot assembly task, we then want to develop methods that allow a programmer to search for useful information for the task at hand in the data base and connect the found elements into a new robot system to facilitate the establishment of new robot solutions.

The session is a follow-up to the International Workshop on Re-Using Robot Data that was held on September 8, 2022 at the University of Southern Denmark ( Approx 100 participants from academia and industry participated in this event that
was supported by DIREC and DDSA.

At the workshop it became clear that there is a strong interest in the
different communities for how data science, digitalization and robotics can join forces. In addition, there is a strong pull from the robotic industry, which struggles with efficiently gathering and exploiting data connected to the set-up of robot systems as well as the running of those systems. 

Target Audience and Size

We want to attract researchers within data science, digitalization and robotics as well as companies. We
aim for 25-30 participants that will discuss the challenges and opportunities. 


  • Prof. Norbert Krüger, SDU | Introduction and moderation | 5 mins.
  • Prof. Mikkel Baun Kjærgaard, SDU | Engineering of AI systems for Automation and Robotics | 10 mins.
  • Assoc. Prof Aljaz Kramberger, SDU | Exploitation of robotic data in real and simulated environments for assembly of pharmaceutical devices | 10 mins.
  • Assist. Prof. Mina Alipour, SDU | User interface challenges for human-robot interaction | 10 mins.
  • Assist. Prof. Hang Yin, KU | Leveraging Simulation in Learning-based Robotics | 10 mins.
  • Assoc. Prof. Andres Masegosa, AAU | Transfer Learning and Applications to Robotics | 10 mins.
  • Panel discussion on “The road ahead?” | Moderator: Prof. Norbert Krüger, SDU and panel: Prof. Mikkel Baun Kjærgaard, SDU, Assoc. Prof Aljaz Kramberger, SDU, Assist. Prof. Mina Alipour, SDU, Assist. Prof. Hang Yin, KU, Assoc. Prof. Andres Masegosa, AAU, Søren Adamsen Mouritzen, Odense Robotics | 25 mins.


  • Norbert Krüger, Professor in Robotics, Mærsk Mc-Kinney Møller Institute, SDU
  • Mina Alipour, Assistant Professor in HCI, Mærsk Mc-Kinney Møller Institute, SDU
  • Andres Masegosa, Associate Professor in Machine Learning, Department of Computer Science, AAU
  • Hang Yin, Assistant Professor in Computer Graphics, Department of Computer Science, KU


Modern organizations face significant transformation brought by the establishment of disruptive technologies. These technologies are the cause for the re-examination of fundamental aspects of today’s organizations, such as the way they operate, create value, communicate but also the way they are governed. Notable technological advancements, such as Artificial Intelligence, Internet of Things, Robots, Drones and Quantum Computing have occupied the agendas of many countries, public authorities, research and education institutions, non-profit organizations, and private enterprises. However, their disruptive nature questions the established foresight studies methods that academics, policy makers and industrial stakeholders rely on in order to be prepared for the future developments of technologies.

We propose the organization of an interactive workshop that will bring researchers and practitioners together, in order to create an avenue for discussions on how such technologies will affect the organizations of the future. The outcomes of the workshop, as discussion points, are expected to assist the knowledge we have on how to study the future of such technologies and to be lenses through which future analyses may emerge. Aligned with the research scope of DIREC, which is leading in hosting research in relevant subjects, as observed in multiple of its research areas (for instance, WS2, WS6, SIA1), we have chosen to orient our topic into four technological

  • Robots and Drones
  • Quantum Computing
  • Artificial Intelligence
  • Internet of Things

Main Activities

The workshop follows the method of World Café (TM). There will be a short introduction to the discussions taking place as well as the general guidelines for the orchestration of the process. Afterwards, the participants will be split into four groups, visiting one table each. In each table there will be one organizer who is going to act as a
moderator for the discussions. The groups will discuss a topic at each table for approximately 15 minutes. The moderator will keep notes which will be presented at the end of the activity. After 15 minutes each group will rotate to the next table clockwise, to discuss another topic. The process will repeat itself until the groups have visited all four tables. Then, in the remaining time, the moderators will present the results of the activity, sparking dialogue among the participants and providing takeaways for them.

The topics discussed in each of the four tables correspond to four disruptive technologies, namely: Robots and Drones, Quantum Computing, Artificial Intelligence, and Internet of Things. Each one of the organizers is conducting research on one of those subjects, so they can effectively moderate the discussion, bringing their insights while promoting the exchange of ideas. The topics discussed in each table correspond to the future effects brought to organizations by each of these technologies, with the discussion unfolding various aspects of these effects, touching upon subjects like the value creation, the challenges, the governance change but also the ethical implications related to these technologies.

The World Café (TM is an established but also interactive method of non-formal education that has been tested in both research (Steier et al., 2015) and business and entrepreneurial contexts (Chang, 2017; Decker-Lange et al., 2021). It assists participants in their effort to engage in constructive dialogue and build collaborations and personal
relationships (Tan & Brown, 2005). In addition, it has already been tested in foresight initiatives (Hines & Whittington, 2017; Štefanić & Šimić, 2021). As a result, it would be a good fit for the overall theme of the conference, which aims to bring the research community together and form collaborations. Since the discussions relate to multiple concepts, this format can lead to comparable discussion outcomes for the four technologies. 

Target Audience and Size

The target audience for this event is diverse. We are interested in the insights emerged from academia, industry, public sector, but also from the civil society, all related to the organizational effects of disruptive technologies. The participants may be interested in the technical aspects of the technologies, or they may be interested in their societal and managerial implications. The multifacetedness of the incorporation of these technologies in modern organizations unfolds the professionals’ and academics’ diverse viewpoints. 

We are especially interested in attracting junior researchers, such as ourselves, in order to foster research dissemination, to inform each other’s research based on the insights emerging from each one’s research initiatives. Moreover, the junior researchers can benefit from meeting with industrial stakeholders participating in the workshop, since they can be informed about their viewpoints on the subject, thus enhancing their understanding of the subject in a more holistic way. Finally, they can benefit from the ability to meet other researchers and stakeholders, since it will fuel collaborative initiatives in the areas of disruptive technologies. The industrial stakeholders may benefit from the participation, since they can be updated on the research on the future of the disruptive technologies, gaining input that can incorporate in their professional work.

We want everyone to have an equal opportunity to express their viewpoint on the subject discussed on each table. As a result, we would like to limit the number of participants to 20. In addition, we would like to pre-arrange the four participants groups, in order to ensure the diversity of backgrounds of the participants in each group, which would magnify the fruitfulness of the discussions. For that, we would like to request the participants to declare their participation through an online form a few days before the workshop.


The organizers of this event, which will subsequently be the moderators of each table are proposed to be the following:

  • Alexandra Hettich, PhD fellow, Department of Digitalization, Copenhagen Business School
  • Ignacio Godoy Descazeaux, PhD fellow, Department of Digitalization, Copenhagen Business School
  • Panagiotis Keramidis, PhD fellow, Department of Digitalization, Copenhagen Business School & DCR Solutions
  • ShiTing Wang, PhD fellow, Department of Digitalization, Copenhagen Business School

We are a group of junior researchers with a diverse range of  backgrounds and research interests, which coincide on investigating the organizational effects of disruptive technologies. Alexandra is conducting research on the application of drones and robots in agriculture, Ignacio is exploring the impact of quantum computing in the management of information systems, Panagiotis investigates the business value of Artificial Intelligence and the various stakeholders’ perceptions related to that, while ShiTing Wang is focused on Cybersecurity and Internet of Things (IoT). The projects of Alexandra, Panagiotis and ShiTing are part of DIREC.

This workshop, being conducted in consideration of our academic citizenship obligations, serves as an opportunity for us to raise awareness on the decisiveness of the impact that disruptive technologies are expected to have on organizations, which will be demonstrated through the dialogue amongst the participants. It is also a great opportunity for us to sense the perceptions and opinions of other researchers and stakeholders on these disruptive
technologies and incorporate the fruitful dialogue results in our own research endeavors.


  • Chang, W. L. (2017). Online training for business plan writing through the World Café method: the roles of leadership and trust. Universal Access in the Information Society, 16, 313-324.
  • Decker-Lange, C., Lange, K., Dhaliwal, S., & Walmsley, A. (2021). Exploring entrepreneurship education effectiveness at British universities–an application of the World Café method. Entrepreneurship Education and Pedagogy, 5(1), 113-136.
  • Hines, A., & Whittington, A. (2017). Nine emerging student needs. On the Horizon, 25(3), 181-189.
  • Štefanić, J., & Šimić, D. (2021). An overview of skills foresight methods. In Central European Conference on Information and Intelligent Systems (pp. 283-289). Faculty of Organization and Informatics Varazdin.
  • Steier, F., Brown, J., & Mesquita da Silva, F. (2015). The World Café in action research settings. The SAGE handbook of action research, 3, 211-219.
  • Tan, S., & Brown, J. (2005). The world café in Singapore: Creating a learning culture through dialogue. The journal of applied behavioral science, 41(1), 83-90


In a world where data can be used to both gain valuable insights into significant problems and to intrude on individuals’ right to privacy, it is increasingly important to investigate means of benefitting from our vast amounts of data without undermining personal privacy. This
issue is specially present in data science applications where user data can be leveraged to benefit end users and larger organizations but also misused to glean sensitive information that these entities did not intend to make public. Privacy preserving computation is a powerful tool for processing data from multiple sources while only revealing the output of such computation and ensuring that potentially sensitive inputs remain private. The goal of this session is to
introduce the main concepts and tools in privacy preserving computation with a focus on how they can be used as tools for enabling collaborative data science activities across organizations.

We intend to introduce attendants to two sets of complementary techniques that have found several applications in regulatory compliant collaborative data processing: Multiparty Computation (MPC) and Differential Privacy (DP). In MPC, mutually distrustful entities collaborate to compute a function of their private inputs while revealing nothing more than the output. However, the output itself may end up leaking sensitive information about the inputs that
can be misused to single out individual entities whose data fits certain patterns. This issue can be solved by DP, which ensures that computation outputs do not leak sensitive information about individual inputs. We will present the basics of these techniques and how they can enable collaborative data science.

Main Activities

This session will start with introductory talks presenting the main tools for privacy preserving computation and talks focusing on how these tools can be efficiently applied to solve important problems in data science. Next, we will hold panel discussions with experts in
this field aiming at identifying new problems for which privacy preserving computation could provide novel solutions. The audience will be encouraged to actively join these discussions and present problems from their own professional backgrounds and research fields that could benefit from privacy preserving computation.

  • Introduction to MPC | Luisa Siniscalchi, DTU | 15 mins.
  • Introduction to DP | Rasmus Pagh, KU | 15 mins.
  • MPC deployments | Carsten Baum, DTU | 20 mins.
  • Practical use cases of DP | Christian Janos Lebeda, ITU | 20 mins.
  • Panel: Where will PPC go? Where do you want it to go? | All Speakers and Organizers | 20 mins.

Target Audience 

This session targets a diverse audience of industry professionals and
academic researchers who wish to learn more about privacy preserving computation and how it can impact their own professional and academic activities. Towards industry professionals, we aim at creating awareness about how these techniques may solve current privacy issues in their industries as well as how they may enable new business. We also aim at introducing academic researchers from different backgrounds to this field so that they may incorporate it in their research and spark new interdisciplinary research projects. 


  • Bernardo David, ITU
  • Carsten Baum, DTU
  • Rasmus Pagh, KU
  • Claudio Orlandi, AU
  • Luisa Siniscalchi, DTU
  • Peter Scholl, AU

ROOM: Salon 20+21

Advanced machine learning and data analytic techniques have demonstrated potential for application in mental health care and psychiatry. Further advancement in sensor technology (camera, microphones, biosensors) has led to efficient data collection mechanisms, that can be leveraged for improved management of psychiatric disorders. Different data sources like audio-visual data, text, central and peripheral autonomic nervous system, and even health registry data are utilised at various stages from screening to diagnosis, post hoc analysis and for management and intervention of psychiatric disorders. Given the growing need for resources in mental health care, this is a positive development as mental health experts can better allocate resources towards patients with the aid of data-driven diagnostics and treatment.

A general goal of machine learning models is to have high generalizability over all data from the target application. However, in practice, the generalization of models deteriorates as the distance between the training and test/deployment data increases. In (mental) healthcare research, variability in the dataset can be high, whereby generalization may not be possible due to insufficient data relative to the amount of variability and small cohort sizes. Furthermore, variability between multiple data cohorts can also be due to differences in gender, culture, socioeconomic differences.

In addition, treatment and management of psychiatric disorders are often require a tailored approach for treatment that is based on the individual or individual differences. These technical challenges raise structural questions regarding the safety and trustworthiness of these systems and how to mitigate the generated risks and potential biases perpetuated through the systems. Given the growing interest and exploration of data-driven technologies for mental health, the aim of this session is to present a balanced view on the use of machine learning in mental healthcare, with a discussion on the opportunities, challenges and limitations. The specific goals are:

  1. Gain and consolidate insights on recent advances and existing unexplored dimensions for the use of machine learning in preventive mental healthcare.
  2. Discuss the balance in human-machine cooperation and how to define the roles of human-experts and machines within this cooperation.
  3. Generate a constructive discussion on the limitations of data-driven approaches within mental healthcare, including ethical and safety considerations, also in light of the EU AI-Act and growing regulation worldwide.

To enable the above session goals and obtain cross-disciplinary insights, we have invited stakeholders with expertise in the following disciplines: psychiatry, machine learning and AI regulation

Session Format

We will organise the special session in the form of 1. Invited lightning talks (15 minutes each), 2. Panel discussion (45 minutes). Within the 90 minutes session, the first 45 minutes is allocated for presentations from the invited experts to set the stage. Following this, next 45 minutes is allocated to an open panel discussion, which will begin with questions to the invited experts and then open the floor for audience engagement. The proposed session will address the advances and limitations of data driven approaches to preventive mental healthcare. We have invited interdisciplinary participants with primary expertise in psychiatry, machine learning and regulatory landscape on human rights in AI. The invited speakers (also panelists) are:

  • Michael Benrose, Head of Research on Biological and Precision Psychiatry Copenhagen Research Centre for Mental Health, Mental Health Centre Copenhagen, Copenhagen University Hospital
    Perspective: Opportunities and potential of using ML in mental health.
  • Line Clemmensen, Associate Professor, Department of Applied Mathematics and Computer Science, DTU
    Perspective: Challenges of using ML in mental health.
  • Cathrine Bloch Veiberg, Chief Advisor, Human Rights, Business & Tech The Danish Institute for Human Rights.
    Perspective: Roles and responsibilities of developers and deployers and the regulatory landscape of requirements on human rights in AI.
  • Panel Moderator: Nicole Nadine Lønfeldt, Senior Scientist, Børne- og Ungdomspsykiatrisk Cente

Target Audience

The target audience is any participant with an interest in or working within ML for (mental) healthcare or other high-impact AI applications. Also, conference participants engaged in applications of data science in disciplines like human-centric computing, social-signal processing, trustworthy AI and related fields may find this session useful. 

We expect an audience of around 50–75 people in the lecture room.


  • Sebastian Basterrech, Postdoctoral research fellow, DTU Compute (STAT),

  • Sneha Das, Assistant Professor, DTU Compute (STAT),

  • Nicole Lønfeldt, Senior Scientist, Child and Adolescent Mental Health Center, Copenhagen University Hospital—Mental Health Services

  • Line H. Clemmensen, Associate Professor, DTU Compute (STAT)



The “Software Engineering in the Age of AI” session aims to discuss trending topics that concern practitioners and researchers in the fields of Artificial Intelligence for Software Engineering (AI4SE) and Software Engineering for Artificial Intelligence (SE4AI). This session covers three major topics: AI code generation, AI enabled program synthesis, and software documentation of AI models & datasets. We combine knowledge derived from practical experience and the state of the art to raise awareness on key aspects that require attention from Danish professionals to support high-quality software development.

Session format and program

Welcome | Thiago Rocha Silva, Associate Professor, SDU (Host) | 5 min. 

Short talk 1 | Mircea Lungu, Associate Professor, ITU | 10 min., Q&A 5 min. 

  • Developer Experience with AI code generation: The good, the bad, and the ugly: This talk covers current trends in AI and what might they mean for software engineering. We will discuss AI code generation from the point of view of the developer: The impact of the state-of-art AI tools on their practice. 

Short talk 2 | Boris Düdder, Associate Professor, KU | 10 min., Q&A 5 min. 

  • Type-based program synthesis and feedback refinement using generative AI: This talk discusses program synthesis, which is used to discover executable programs from the user intent expressed in the form of constraints. We will focus on the program synthesis supported by generative AI, which has been largely used to generate high-quality text, images, and other types of content. 

Short talk 3 | Eduardo Fernandes, Asssitant Professor, SDU | 10 min., Q&A 5 min. 

  • On documenting AI models and datasets: This talk discusses the importance of documenting AI models and datasets, especially in online data stores such as Hugging Face. We will discuss existing guidelines, current practices and limitations of such documentation.  

Discussion panel | Thiago Rocha Silva (moderator) | 45 mins. 

  • Boris Düdder
  • Eduardo Fernandes
  • Mircea Lungu

Target audience and size

The “Software Engineering in the Age of AI” session targets practitioners concerned with either Software Engineering practices in general or AI4SE/SE4AI practices specifically, as well as researchers interested in understanding current trends and relevant industry challenges that could be addressed in future work. We expect a target audience of 20-30 participants.

ROOM: Salon 31

This session aims at bringing together the danish algorithms researchers community to share the latest developments. The aim is in particular, to showcase new PhD students, PostDocs, young faculty and their projects, discuss new research projects (already funded or at the stage of ideas), and show recent breakthrough results.

Main Activities

  • Joan Boar, Lene: Paging with Succinct Predictions.
  • Shrikant, Amik Raj Behera: Sublinear-time algorithms for decoding some error-correcting codes.
  • Mingmou, BARC: Sparsity-Dimension Trade-Offs for Oblivious Subspace Embeddings
  • Shuo Pang, Jakob Nordström: Graph Colouring Is Hard on Average for Polynomial Calculus and Nullstellensatz

Target Audience

All active algorithms researchers in Denmark, specifically across seniority and from the different institutions. Expected 40 participants. No cap.


  • Riko Jacob, IT University of Copenhagen
  • Kasper Green Larsen, Aarhus University

ROOM: Salon 26+27

Data science is on everybody’s mind and we see an increasing curiosity from both the media and the general public. New scientific discoveries that would previously only have been of interest to a small community are now broadly communicated and discussed in mainstream media. This overwhelming interest in everything “data science”, “AI” and “machine learning” has been wonderful for the community, but it also comes with a we also see a downside: Inaccurate, implausible and out-right wrong ideas about data science, statistics and uncertainty being spread in the public – often by well-meaning journalists. This shows a need to facilitate a more well-informed public debate about these topics, and that calls for more active science communicators from the data sciences!

The purpose of this session is to start a conversation between data scientists and journalists in order to break down communication barriers from both sides of the table, and enable a more visible role of data scientists in mainstream media. We will have speakers from the press, as well as some of the few data scientists that already have experience with helping journalists, so we can get multiple angles on what it takes to further facilitate a more active role of data science in public media.

The session thereby helps data scientists that are interested in communicating research through the media or in engaging in public discussions, by demystifying the journey from having something to say, to actually being out there, talking to the press and the general public. By taking this step we hope to narrow the gap between data scientists, the press, and ultimately also the public, helping all to better understand each other.


The session will consist of four brief presentations (each about 10 minutes). Two will be from data scientists and two will be from journalists. They will each provide insights into their personal experiences with working together with the “others” – media or scientists – and reflect on the challenges they have met, as well as the rewards they have gained from talking to each other.

Afterwards, the speakers will all participate in a moderated panel discussion (about 45 minutes), focusing on identifying and breaking down barriers for participating in the public debate. There will be amble opportunities for the audience to participate by providing questions and input for the panel to discuss.

Target Audience

Anyone with an interest in public science communication and public media. We welcome participants both with and without personal experience with science communication and press contact.


Confirmed speakers:

  • Susanne Ditlevsen, Professor of Statistics, University of Copenhagen
  • Anders Høeg Lammers, Journalist and Editor for Forskerzonen at
  • Leon Derczynski, Associate Professor of Computer Science, ITU
  • Mikkel Bødker Olsen, Jounalist and Opinion Editor at Børsen

Chair: Anne Helby Petersen, Assistant Professor of Statistics, University of Copenhagen and board member of The Danish Statistical Society

Panel moderator: Theis Lange, Professor of Statistics and Head of Department, Univesity of Copenhagen


This session is organized by the Danish Statistical Society’s committee for Communication and Press (DSTS-C&P).

  • Anne Helby Petersen, Assistant Professor, Section of Biostatistics, University of Copenhagen & board member of the Danish Statistical Society and DSTS-C&P.
  • Christoffer Sejling, PhD Student, Section of Biostatistics, University of Copenhagen & member of DSTS-C&P.
  • Michael Sachs, Associate Professor, Section of Biostatistics, University of Copenhagen & member of DSTS-C&P.
  • Theis Lange, Head of Department and Professor of Biostatistics,
    Department of Public Health, University of Copenhagen & member of DSTS-C&P

ROOM: Sal C + Vandrehal mod park

Dedicated time for social activities to foster connections among participants, 

We encourage you to bring along your current knitting project or (transportable) musical instrument, providing an opportunity to sit together and share your passions while getting to know fellow participants better.

Additionally, we have arranged for a selection of fun board games and interactive challenges involving small robots, ensuring a playful and enjoyable atmosphere for everyone. 

See social menu

ROOM: Glassal + Pavillon + Veranda + Terrassen


  • Tartar of baked beetroot, herb salad, raw pickled Jerusalem artichokes, roasted nuts and beluga lentils in vinaigrette.
  • Veal braised in red wine, honey glazed parsnip, fried mushrooms, baked onion, potato pure with watercress and brown butter,  red wine sauce.
  • Lemon cake with burned meringue, vanilla ice cream, caramel sauce, and crunchy candied pumpkin seeds

Vegetarian/vegan menu:

  • Tartar of baked beetroot, herb salad, raw pickled Jerusalem artichokes, roasted nuts and beluga lentils in vinaigrette.

  • Steak of potato, leek and root vegetables, honey-glazed parsnip, roasted mushrooms turned with shallots, baked onion, petite potatoes, vegan red wine sauce.

  • Vegan chocolate mousse with berry compote and candied pumpkin seed crunch.

ROOM: Stjernebaren and Vandrehal mod Park

Dedicated time for social activities to foster connections among participants, 

We encourage you to bring along your current knitting project or (transportable) musical instrument, providing an opportunity to sit together and share your passions while getting to know fellow participants better.

Additionally, we have arranged for a selection of fun board games and interactive challenges involving small robots, ensuring a playful and enjoyable atmosphere for everyone. 

NB! After dinner, attendees will have to pay for their own consumption in the bar.  The bar accepts cash or card.

February 2, 2024

If you want to start the day full of energy, you’ll have the possibility to join a: 

  • running group for a 5K run
  • winter bathing group
  • yoga class (max. 40 participants)

You can sign up for your preferred activity when you arrive at the venue.

If you need some time to relax, the hotel has available a Spa & Fitness area with sauna, hot water spa with jacuzzi and cold-water pool.


When to trust AI…

Machine learning solutions are revolutionising AI, but their instability against adversarial examples – small perturbations to inputs that can catastrophically affect the output – raises concerns about the readiness of this technology for widespread deployment. Using illustrative examples, this lecture will give an overview of techniques being developed to improve the robustness, safety and trust in AI systems.

About Marta Kwiatkowska

Marta Kwiatkowska is Professor of Computing Systems and Fellow of Trinity College, University of Oxford. She is known for fundamental contributions to the theory and practice of model checking for probabilistic systems, and is currently focusing on safety, robustness and fairness of automated decision making in Artificial Intelligence.

She led the development of the PRISM model checker (, which has been adopted in diverse fields, including wireless networks, security, robotics, healthcare and DNA computing, with genuine flaws found and corrected in real-world protocols. Her research has been supported by two ERC Advanced Grants, VERIWARE and FUN2MODEL, EPSRC Programme Grant on Mobile Autonomy and EPSRC Prosperity Partnership FAIR.

Kwiatkowska won the Royal Society Milner Award, the BCS Lovelace Medal and the Van Wijngaarden Award, and received an honorary doctorate from KTH Royal Institute of Technology in Stockholm. She is a Fellow of the Royal Society, Fellow of ACM and Member of Academia Europea.

ROOM: Salon 20+21

The computational need of the powerful ML models has increased five orders of magnitude in the past decade. A state-of-the-art natural language processing model to be deployed can cost hundreds of thousands of euros to train in the cloud without accounting for the electricity cost and carbon footprint. This makes the current rate of increase in model parameters, datasets, and compute budget unsustainable. To achieve more sustainable progress in ML in the future, it is essential to invest in more resource-/energy-/cost-efficient solutions. In this session, we will explore how to make ML’s computational and carbon footprint more transparent and improve ML resource efficiency through software/hardware co-design. We plan to take a holistic view of the ML landscape, which includes data preparation and loading, continual retraining of models in dynamic data environments, compiling ML on specialized hardware accelerators, and serving models for real-time applications with low-latency requirements and constrained resource environments. 

Target Audience

This session aims at reasoning critically about how we build software and hardware for end-to-end machine learning. We hope that the discussions will lead to increased awareness for understanding the utilization of modern hardware and kickstart future developments to minimize hardware underutilization. We are thus expecting to see interest and contribution from academics (especially PhD students) and industry across fields of data management, machine learning, systems, and computer architecture covering expertise of algorithmic optimizations in machine learning, job scheduling and resource management in distributed computing, parallel computing, and data management and processing.

Main Activities

The overview of our 90 mins. session is as follows:

  • Brief introduction of the session | 10 mins.
  • Talks (~7 mins. talk, ~3 mins. Q&A) from people approaching the challenge of sustainability of ML from different angles | 60 mins.
  • Discussion hour with preprepared questions by the organizers and comments questions from the audience. The questions we don’t have the time to answer in the first hour can
    also be covered in this hour | 20 mins. 


  • Ingrid Munne, Data Scientist and ML Engineer at Electricity Maps
  • Robin Troesch, Data Engineer at Electricity Maps
  • Pedram Bakhtiarifard, Research Assistant at University of Copenhagen and maintainer of the Carbontracker framework
  • Ehsan Yousefzadeh-Asl-Miandoab, PhD student at IT University of Copenhagen
  • Mahyar Tourchi Moghaddam, Assistant Professor at University of Southern Denmark
  • Julie Koefoed Bielefeldt, Business Developer for Datahubben at Energinet


  • Pınar Tözün, Associate Professor, IT University in Copenhagen
  • Raghavendra Selvan, Assistant Professor, University of Copenhagen
  • Daniel Hershcovich, Assistant Professor, University of Copenhagen
  • Ties Robroek, PhD student, IT University in Copenhagen
  • Julian Schön, PhD Student, University of Copenhagen


ROOM: Salon 25

While many are hopeful about the advent of quantum computing and its promises of solving many computational problems more effectively than we can do today, cryptographers and IT-security experts are more worried about the ability of quantum computers to break most of the public-key cryptographic algorithms which are currently deployed and securing our digital interactions.

This session will provide some introductory talk to the topic of cryptography and IT-security in a quantum world, trying to separate the hype from the hope.

Main Activities

The list of speakers and titles/topics is as follows:

  • Why quantum computing is different from classical, and how can quantum computers break current cryptography? | Damgård | 25 mins.
  • Challenges in proving security against quantum computers | Majenz | 25 mins.
  • Secure cryptographic primitives in a quantum world, including Danish submissions to the international NIST post-quantum competition | Baum/Scholl | 25 mins.
  • Q&A | 15 mins.

Target Audience

Anyone interested in cryptography, IT-security, and quantum computing. The talks will be tutorial style for a large computer science audience. We do not have any restriction on the number of participants. 


The session is co-organized by the speakers, and the chair and co-chair of the DIREC Cybersecurity Workstream who will host the session.

  • Carsten Baum, DTU
  • Ivan Damgård, Aarhus University
  • Christian Majenz, DTU
  • Claudio Orlandi, Aarhus University
  • Peter Scholl, Aarhus University


Equity, that is, freedom from bias and favouritism, in non-diverse settings such as many parts of academia, is not easily obtained. There are subtle structures which hinder equal opportunity, but unfortunately also harassment cases. Whether you are a junior researcher, a colleague, a supervisor, a personnel manager, or an organiser of a conference, workshop or event, cases of discrimination or harassment are difficult to handle:

  • If you are on the receiving end of discrimination or harassment, reacting comes with a risk of repercussion;
  • If you witness discrimination or harassment, you likely do not know what to do, and the target will often ask you not to do anything;
  • Even if you try to prepare yourself, for example, by endorsing a code of conduct for your event, it does not prevent harassment, and your ability to react is severely limited by the target’s willingness to come forward;
  • Often perpetrators have never been on the receiving end of discrimination and harassment, and have never been confronted with their behaviour being inappropriate. This leads to an automatic defensive response which hinders structural change and understanding;
  • In essence, we have very few tools at our hands, and they are more dedicated to supporting victims after the fact, than towards avoiding discrimination and harassment in the first place. Outside of formal legal proceedings involving the police and HR, there are very few possibilities for us as a community to address this issue. In particular, aside from introducing a code of conduct for an event/organisation, it is almost impossible to institute long term consequences for repeat offenders, and any possible action tends to be reactive, rather than proactive. As such, it is important to ensure that the Danish data science and digitalisation communities commit to having open discussions about expectations of behaviour, and work towards a sustainable long term framework to minimise harassment and discrimination.

In this session, we hope to create awareness that discrimination and harassment persist; create a new discussion on how to improve this aspect of our scientific and technical culture; and refresh the discussion on what a good code of conduct looks like – and how to make it count.


  • Introduction | 10 mins.
  • Talk by Maria Mortensen, EDI-Consult | 25 mins.
  • Facilitated group discussion. Discussion cards describing cases with themes such as: How to handle an unacceptable situation – as a bystander, as an organiser. How to handle a maybe unacceptable situation – as a bystander, as an organiser. Why is it hard to recognize such situations? What can the system do? Should there be repercussions and if
    so, what should they be? | 50 mins.
  • Plenum – gathering our thoughts and ideas | 5-10 mins. 

Target Audience

Everyone interested in changing the culture. 50-60 people. 


  • Lisbeth Fajstrup, AAU, DDSA Education and Networking Committee, ENC
  • Aasa Feragen, DTU, DDSA Education and Networking Committee, ENC
  • Sebastian Weichwald, DDSA Young Academy Panel, YAP
  • Yossi Bokor Bleile, AAU


Dive into the world of algorithmic fairness within healthcare. This
session provides an opportunity to participate in a hands-on hackathon where you’ll work with real-world data and algorithmic biases. By delving into the ethical and societal implications of your work, you’ll not only enhance your data science skills but also contribute to a future where responsible AI is the norm. You will receive a starter toolkit so that you can begin to consider fairness in your work with algorithms. Join us to address healthcare disparities and make a meaningful impact.

What is going to happen?

Participants will engage in a session featuring a brief presentation on algorithmic fairness and bias examples in healthcare, followed by a hackathon with group work. The event will conclude with discussions and the institutionalizing of the results.

Target Audience

Join us if you’re into algorithmic fairness! Open to PhD students, postdocs, professors, data experts, and related professionals. Must have intermediate to advanced coding skills in R/Python. 


  • Tibor V. Varga, Associate Professor, Section of Epidemiology, Department of Public Health, KU
  • Adrian G. Zucco, Postdoc, Section of Epidemiology, Department of Public Health, KU

ROOM: Salon 30

Process Mining refers to algorithms for learning process descriptions from event logs. An event log is a collection of ordered events recording steps of the execution of a process, which could be a business process in a company, a case management process in a municipality or a production process in a factory.

Process mining has many uses, includig Discovery of processes, Performance analysis, Root cause analysis, Predictive analytics, Compliance and Auditing, Continuous Improvement, Decision support.

DIREC hosts several researchers who are active contributers to the process mining community and will be hosting the International Conference on Process Mining (ICPM 2024) in the Fall 2024 at DTU (

The process mining field has several touch points with other areas of research in DIREC, which is indicated by the usual co-located workshops at the ICPM conference, that range from Responsible process mining (RPM), Streaming Analytics for process mining (SA4PM), Process-oriented Data Science for Healthcare (PODS4H), Process-querying, manipulation and intelligence, Leveraging Machine Learning in Process Mining (ML4PM), Education meets
process mining (EduPM). 

Main activities

  • Welcome and brief presentation of participants | Thomas Hildebrandt | 10 mins.
  • Process mining research at DTU: Andrea Burattin & Hugo A Lopez | 20 mins.
  • Industrial process mining tool (Apromore) demo: Breakaway | 20 mins.
  • Process mining research at DIKU: Tijs Slaats, Axel K F Christfort, Paul Cosma | 20 mins.
  • Industrial process mining tool (DisCoveR): DCR Solutions | Morten Marquard | 20 mins.
  • Plenum discussion on the road ahead and collaboration | 20 mins. 

Target Audience

The target audience is on the one hand researchers and practitioners in process mining and on the other hand researchers and practitioners in areas that could contribute to process mining research (Machine Learning, Time Series Management, Streaming Analytics, Behavioural Querying, Data Quality, and Governance) or where process mining could be applied (healthcare,
IoT, Education, eGovernment …)


  • Andrea Burattin, DTU
  • Hugo Lopez, DTU
  • Tijs Slaats, KU
  • Axel Christfort, KU
  • Thomas T. Hildebrandt, KU

ROOM: Auditorium

Machine learning (ML) theory provides the foundational understanding of AI methodology. Despite much activity in the Danish research community with many contributions at the major ML venues, there have not been any regular meetings in Denmark bringing the ML theory community together.

The goal of this session is to change that and to provide a forum to share ideas and explore potential synergies in theoretical ML.

Main activities

  • Welcome | 5 mins.
  • Talk 1 | 20 mins.
  • Talk 2 | 20 mins.
  • Talk 3 | 15 mins.
  • Panel discussion: How do we strengthen the machine learning theory
    community? What are the major open theoretical questions in machine learning and how will their solution impact machine learning and artificial intelligence | 30 mins. 

Preliminary list of speakers (to be extended)

  • Melih Kandemir, SDU
  • Arijit Khan, AAU
  • Jun Yang, KU

Target Audience

The target audiences are researchers in Denmark who are interested in machine learning theory.


  • Ole Winther, DTU and KU
  • Christian Igel, KU
  • Yi-Shan WU, SDU

ROOM: Salon 28+29

Entrepreneurship is a central aim of Danish universities and several initiatives are initiated to support research commercialisation, start-ups, research collaborations and other types of entrepreneurship. At this workshop we will address how researchers who engage in entrepreneurial activities can reap benefits for their research, teaching and academic careers, and on how they can be supported in this.

The background of this workshop is that DIREC (WP13) in collaboration with DTU Entrepreneurship has initiated a project on university based or academic entrepreneurship within the digital area. The project addresses key the following themes:

  • Factors that increase the likelihood of academic entrepreneurship – what can institutions do to foster academic entrepreneurhip
  • Effects of academic entrepreneurship on individual researchers & research – the role and effects of entrepreneurship on teaching and training

The aim of the project is to gather and disseminate knowledge about key factors and best practices in promoting digital academic entrepreneurship, more specifically to get:

  • Insight into factors that support or hinder digital entrepreneurship
  • Increased entrepreneurial activity
  • Strengthened benefits for entrepreneurial researchers
  • Inputs to strategy processes in universities and departments

The project’s deliverables and elements are:

  • White paper on how to foster (digital) academic entrepreneurship – based on a review of the literature
  • Workshop with DIREC partners (November 2023)
  • Interactive session at D3A conference (February 2024)
  • 4-6 illustrative cases of digital academic entepreneurship
  • Digital Entrepreneurship guide summarizing key findings for university managers, VIPs and innovation support staff etc.

This workshop at D3A aims to discuss the key results the this project, in particular the draft of the White Paper and the Digital Entrepreneurship Guide, and to gather participants’ insights and good practices for promoting digital academic entrepreneurship. Your participation will provide valuable inputs to the development of a guide to supporting digital entrepreneurship for researchers and university managers. Please participate and use your experience to support others to be successful in digital entrepreneurship.


  • See slides
  • Presentation: ”Entrepreneurship guide – key results and lessons learnt” | Maria Theresa Norn, Associate Professor, Centre for Technology Entrepreneurship, DTU and Senior Researcher, Department of Political Science – Danish Centre for Studies in Research and Research Policy, Aarhus University | 30 mins. 
  • Presentation: “Entrepreneurship in practice at universities – tools and support” | Christian Dalsgaard Nielsen, Innovation Partner, DTU Skylab | 15 mins. 
  • Break | 10 mins. 
  • Presentation of 2 university based startup cases | 10 mins. 
    • Marius Cortsen, CEO and Co-Founder, ChatTutor
    • Line Clemmesen, CSO and Co-founder,
  • Panel discussion and debate | Towards The Entrepreneurial University | 20 mins. 
    • Mads Nielsen, Professor, DIKU
    • Line Clemmensen, CSO and Co-Founder,
    • Thomas Riisgaard Hansen, Director, DIREC
    • Søren Poulsen, Senior Consultant, AU CS
    • Marius Cortsen, CEO and Co-Founder, ChatTutor
  • End of workshop: Wrap up and next steps | 5 mins. 

Target Audience

Entrepreneurial researchers/PhD-students, Business developers, tech transfer officers, and university based pre-startups/startups. 


  • Mark Riis, Head of Innovation, DTU Compute
  • Anders Pall Skött, Head of Business and Innovation, DIKU (moderator)

ROOM: Salon 26+27


The modelling of environmental systems requires a very strong understanding of the underlying processes and stakeholder needs. Without this insight, it is near impossible to develop meaningful data-driven approaches. On the other hand, environmental modelers still have limited understanding of the potentials and best practices of employing machine learning techniques. The potential of e.g. integrating physical system understanding in machine learning, or the efficient inverse identification of unknown system parameters remains largely unexplored. Our aim with this workshop is to gather Danish best practice examples for applying data science in environmental applications, as well as to foster dialogue both within environmental sciences and across environmental and computer science. As part of the session, we will discuss the framing of a network on environmental data science

Target Audience

Young and mid-level researchers with environmental background and strong interest in data[1]driven approaches, or from computer science with an interest in environmental applications.

Session format

We expect up to 30 participants. The session will follow a PICO format:

  • 30 min. | Pitch presentations from the presenters, 2-3 min. each
  • 40 min. | Presenters are at their posters and discuss with interested audience
  • 20 min. | Wrap-up + plenum discussion on the potential network on environment data science  


The presentations will cover the development and application of data-driven techniques in an environmental context:

  • Phillip Aarestrup (DMI) – Beyond Rating Curves: Predictions of River Flows and Water Levels with Automated Differentiable Models
  • Anders Bjørn Møller (Aarhus University) – High-resolution soil texture maps based on satellite products
  • Anton Jakob Sørensen (DTU) – Dynamic emulation of 2D flood simulations using scientific machine learning
  • Tanja Denager (GEUS) – Using high resolution satellite data and machine learning to track surface water saturation frequency of Danish peatlands
  • Jun Liu (GEUS) – CAMELS-DK: An Extensive Repository of Hydrometeorological Time Series and Catchment Attributes for 3351 Basins in Denmark for machine learning advancements
  • Amelie Beucher (Aarhus University) – Assessing Danish peatlands over the last decade: mapping and interpreting the spatiotemporal variations of soil organic carbon stocks
  • Sheng Wang (Aarhus University) – Quantifying Climate-smart Agriculture Practices in Denmark and Europe Using Cross-scale Sensing and Process-guided Machine Learning
  • Jonas Wied Pedersen (DMI) – Towards a new database of historical flood events (1990-2020) in Denmark based on newspaper reportings


  • Roland Löwe, DTU 
  • Anders Bjørn Møller, AU
  • Julian Koch, GEUS
  • Jonas Wied Pedersen, Danish Meteorological Institute (DMI)

ROOM: Salon 31

Danish society has a robust commitment to the future of energy and a historical tradition of world-leading industry and logistics. Physical systems carry unique constraints on the modelling and learning systems we build for them, in ways that are only now being explored through domain-specific architectures and integrated systems for software and data engineering (see Physics-informed machine learning and Scientific discovery in the age of artificial intelligence). This session champions the excitement and curiosity of current research for machine-learning in physics and physical industrial systems, and the sustaining impact this research holds for Denmark and the EU. This is the first part of two joint parallel sessions, continuing with Geometric Deep Learning for the Sciences. Attendees are warmly encouraged to benefit from the interactive programme that combines topics across these sessions.

Book a lightning introduction talk.

Session Overview

  • Prof. Allan P. Engsig-Karup, DTU | An introduction to the connections from machine-learning to physics through physics-informed neural networks (PINNs) | 20 mins.

  • Nikolas Borrell-Jensen, DTU | In this application from acoustics, Nikolas presents his recent work on sound field predictions with PINNs | 20 mins.

  • Freja Terp Petersen, DTU | Freja presents work from her recent Masters project on Physics-informed neural networks | 20 mins. 

  • We warmly encourage each participant to present a one-minute, one-slide personal introduction to their interests or research. We will solicit material in advance, and identify common themes for discussion in the following session. Please indicate your interest via the session webpage | 30 mins. 

Target Audience

The organizing team prioritizes efforts by early-career researchers and regards a diverse and inclusive conference session as necessary for strengthening Denmark’s data science discipline. Attendees should treat this session as an opportunity to: grow their visibility with exactly the audience of peers who can support their career development; and to see themselves as enablers of Danish physical data science, including presenting project proposal ideas, concepts for future works as well as developing a framing of their work as addressing societal challenges.


  • Berian James, DTU & Pioneer Centre for AI
  • Julija Tastu, A.P. Møller Mærsk
  • Stefan Pollok, DTU Energy


Denmark has a rich history of evidence-based public health decisions, pioneering developments in life science, and a strong tradition in epidemiology and biostatistics. Moreover, it stands out as a globally recognized digitally advanced country, boasting high-quality data, particularly within the health domain.

As the fields of data science, machine learning, and artificial intelligence continue to flourish, Denmark has witnessed numerous initiatives, research groups, and communities dedicated to pushing the boundaries of innovation in these areas.

The proposed session aims to explore the powerful synergy at the intersection of Denmark’s long-standing expertise in epidemiology, its abundant reservoir of high-quality health data, and the dynamic national data science community.

By harnessing these three interconnected components, we can pave the way for groundbreaking advancements in public health. In addition, the session will touch upon the ethical implications of utilizing vast amounts of health-related data, ensuring that the data science community actively addresses and mitigates potential risks.

Key Objectives

  • Showcasing research: We have a diverse panel of keynote speakers to present case studies that highlight the potential and/or actual impact of data science, machine learning, and AI in public health.

  • Fostering Collaborations: By bringing together experts from different domains, we seek to foster collaborations. Collaborative efforts hold the potential to maximize the value of available data by leveraging cutting-edge algorithms that directly benefit the Danish community and beyond.

  • Identifying Future Directions: The session will conclude with a forward-looking discussion, where participants can collectively envision the future of data-driven public health in Denmark.

Main Activities

We propose a workshop-style session that combines presentations and interactive activities to encourage active participation and knowledge exchange.

Introduction | 3 mins.

  • Overview of the session objectives and structure
  • Introduction of the session organisers and invited speakers

Research Overview | 45 mins.

  • Keynote speakers will present overviews of current research and
    developments from different perspectives
  • Each speaker will highlight their respective areas of expertise and discuss their potential solution to societal/clinical challenges and/or related concerns
  • Each speaker will provide a problem/dilemma question related to the perspective of her/his talk.

Brainstorming session | 40 mins.

  • Participants will be divided into groups
  • The session will focus on the specific problems presented encouraging collaboration, brainstorming, and problem-solving (30 minutes)
  • Each group will share key insights and outcomes from their brainstorming session (10 minutes)

Summary and Closing | 2 mins.

  • The session organizers will summarize the key takeaways and highlight potential avenues for future collaborations

Target Audience

Researchers and practitioners within data science, biostatistics, public health, epidemiology, medicine.

Max 30 participants (excluding speakers and organizers)

Confirmed Speakers

  • Carsten Utoft Niemann, Clinical Associate Professor at Department of Clinical Medicine (UCph), Head of CLL Lab at Department of Haematology, Rigshospitalet, BoD at DDSA
  • Melanie Ganz-Benjaminsen, Associate Professor at Department of Computer Science (UCph), Researcher at Neurobiology research unit at Rigshospitalet, CAC at DDSA
  • John Shorter, Assistant Professor of Biomedical Data Science at Department of Science and Environment, Roskilde University
  • Livie Yumeng Li, PhD Student at Steno Diabetes Center (Aarhus) & Adam Hulman, Associate Professor and Senior Data Scientist, Steno Diabetes Center (Aarhus), CAC at DDSA


  • Federica Belmonte, Statens Serum Institut
  • Cecilie Utke Rank, Rigshospitalet/University of Copenhagen
  • Jens Ulrik Hansen, Roskilde University

ROOM: Glassal + Pavillon + Veranda + Terrassen

ROOM: Salon 20+21

Data has become a democratic concern in several respects, as decisions on society-wide issues of concern as well as individual cases are increasingly data-driven. Similarly, cyber security and data protection have become central policy issues. Thus, citizens’ data are both democratic resources and democratic liabilities – to be activated for policy purposes and kept safe for the individual. At the same time, data is at the heart of current technological developments, with growing concerns about the ways in which generative AI systems and other cutting-edge technologies not only rely on data for their training but reproduce the biases of available data sets in their operations. Thus, a plethora of issues regarding the rights of usage of training data as well as the responsible application of trained models to real-life situations are added to existing questions regarding the responsible use of citizens’ data. In a nutshell, how do individual rights to privacy align with the potential benefits of technological developments? And how do we ensure that such developments are, indeed, beneficial to all members of society, safeguarding everyone from potential harms? More specifically, have digital platforms become critical infrastructures of democratic societies? And has continued technological innovation become a public priority? If so, how might we best protect citizens and communities whilst harnessing their data for the greater common good? In short, how might we envision data democracy?


This panel takes the form of an interdisciplinary panel discussion, bringing together researchers from the humanities, the social sciences, and the technical sciences to discuss sociotechnical issues of data democracy, including but not restricted to questions of platform governance, data protection, and ethical uses. Thus, we begin from questions of data, discussing their interrelations with broader issues of the governance of technological developments. The panel will consist of four researchers from different fields who will each hold a 15-minute presentation, laying out their take on the topic. The panel presentations will be followed by a 30- minute Q&A and general discussion with the audience.

The panel consists of:

  • Anja Mie Weile, Head of Tech4Civ section and PhD student, DTU Compute.
    Title of talk: Tech4Civ – How democratizing technologies for people is important if we are to create sustainable solutions for the future.
  • Attila Márton, Associate Professor, Department of Digitalization, Copenhagen Business School.
    Title of talk: Digital Resilience: What can we learn from ecological thinking?
  • Hjalte Betak, Industrial PhD student, Roskilde University/Kooperationen.
    Title of talk: Who should control which data? Cooperative strategies for individual and collective data control and usage in digital platform work.
  • Irina Shklovski, Professor, Department of Computer Science/Department of Communication, University of Copenhagen/TEMA GENUS, Linköping University.
    Title of talk: The perils of visibility: Reframing notions of data and privacy

Target Audience

All researchers interested in interdisciplinary questions of the democratic use and protection of data, regardless of disciplinary starting point. 


  • Sine N. Just, Professor, Department of Communication and Arts, Roskilde University/the Algorithms, Data &


In the research field of Science, Technology and Society studies (STS) (Danholt & Gad, 2021; Sismondo, 2010), it is well established that technologies are not solely rational, neutral, deterministic, or just a matter of technological progress. Rather they are products of complex processes that includes social, technological, political, economic etc. elements and forces. As such, technologies are assemblages of multiple human and other-than-human actors (Haraway, 1990; Latour, 1994). Furthermore, technologies lead to both expected and unexpected consequences: they are ‘wild’ or wicked problems (Buchanan, 1992; Haraway, 1997; Pickering, 1995). Lastly, the more technologically sophisticated, complex and network-like technologies are, the more opaque and potentially risky they become (Edwards, 2003; Latour, 1992; Perrow, 2011)

This places us as users and developers of technology in a different relation to technology. As Michel Serres once noted the only thing that seems to escape our mastery is the consequences of our actions (Serres, 1995). So instead of assuming and reproducing the somewhat simplistic and utopian idea of humans as masters of the world and technologies, we should perhaps begin to develop other and more adequate narratives and imaginaries (Stengers, 2015).

In this panel, we therefore invite participants to share other stories and ways of being with – living with – technological systems, such as digital systems and AI, than as systems we control or ‘ought to’ control since most of us – also the tech. savvy – clearly don’t. We invite the participants to take their guards down and allow themselves to be overwhelmed by the incomprehensibility, monstrosity, uncertainty, and unruliness of technological systems. We invite the participants to present their experiences and concerns without being obligated to also provide clear answers or suggestions for ‘technical fixes’ or re-design. We invite the participants to share their concerns and uncertainties in order to develop different strategies and solutions than technical ones. To break the spell of technological fixes and try to develop other types of solutions such as, ways of thinking or formulating the problem differently or proposing social or cultural fixes or….? In short, we invite the participants to think about technological systems as if they were trolls, dragons or wild animals that you cannot tame or kill, but must tolerate and learn to live with somehow.

Main Activities

  • Welcome and presentation of workshop | 10 mins.
  • Split up in groups of 4-6 persons
    • Sharing experiences in groups | 30 mins. 
    • Break | 10 mins. 
    • Developing 2-5 answers/strategies/modes of thinking and relating to complex technological systems | 20 mins.
    • Sharing of the group’s insights | 20 mins. 

Target Audience

We consider the workshop relevant to people that work with or encounter digital technologies solutions such as AI, ML or other complex systems. The aim of the workshop is to provoke new ways of thinking and imagining and thereby enable the participants to think in novel ways about technological systems. In preparation of the workshop, the participants are encouraged to think about the incomprehensibility, irrational, creepy, opaque or similar qualities of technology.

Max. 30 participants.


  • Peter Danholt, Associate Professor, Dept. of Digital Design and Information Studies, Aarhus University
  • Frederik Vejlin, Postdoc, Dept. of Digital Design and Information Studies, Aarhus University
  • Kasper Schiølin, Assistant Professor, Dept. of Digital Design and Information Studies, Aarhus University
  • Mathias Alegría Rosa Gil, PhD Student, Dept. of Digital Design and Information Studies, Aarhus University
  • Nina Frahm, Postdoc, Dept. of Digital Design and Information Studies, Aarhus University
  • Joakim Juhl, Lecturer, Technical University of Munich

ROOM: Salon 26+27

In a healthcare landscape increasingly influenced by big data, machine learning (ML), and epidemiology, the confluence of these disciplines holds the key to unlocking substantial improvements in patient care and drug safety. This session aims to delve into the transformative potential of integrating data science with epidemiological insights to improve healthcare outcomes and establish robust pharmacovigilance mechanisms.

The session will feature representatives from three groundbreaking projects. Each panelist will present an overview of their project, focusing on how they utilize data-driven methodologies in synergy with epidemiological principles to enhance healthcare decision-making. Case studies, use-cases, and real-world examples will be shared to demonstrate the practical applications and impact of this multidisciplinary approach.

Invited projects/speakers

  • Ass. Prof. Espen Jimenez Solem, Bispebjerg and Frederiksberg Hospital: “PHAIR project – Pharmacovigilance by AI Real-Time Analyses”
  • Ass. Prof. Carsten Utoft Niemann, Rigshospitalet: “Improving cancer outcomes using machine learning and epidemiology.”
  • Ass. Prof. Martin Sillesen, Rigshospitalet: “Optimizing surgical outcomes through AI enchanced data capture and risk prediction”

Post-presentations, the panel will engage in a moderated discussion, open for audience participation, to debate challenges, best practices, and future directions. The session aims to foster a collaborative atmosphere, bringing together experts, academicians, and practitioners from the fields of ML, epidemiology, and pharmacology. This interdisciplinary dialogue will encourage new collaborations and contribute meaningful advancements to each participating domain.

Main Activities

The 90-minutes session will feature three presentations, followed by a panel discussion. There will also be dedicated time for audience questions.

Session Moderator: Professor Mads Nielsen, PI in the PHAIR project and P1 Collaboratory Co-Lead.

  • Introduction: Welcome and session overview by the moderators | Mads Nielsen and Jesper Kjær | 5 mins.
  • Presentations: Three speakers (20 minutes each) cover topics in Data-Driven Approaches to Healthcare Outcomes and Prediction models | 60 mins.
  • Interactive Discussion: Moderator-led discussion and audience questions | 20 mins. 
  • Closing Remarks | Moderator summarizes key points | 5 mins. 


  • Mads Nielsen, Professor, PhD, Department of Computer Science, KU
  • Jesper Kjær, Director of Danish Medicines Agency’s Data Analytics Centre, Co-chair HMA, EMA Big Data Steering Group, Danish Medicines Agency 

ROOM: Salon 25

Recognizing the interest in AI and data’s organizational aspects, which are often viewed from a data science angle, our session targets a cross-disciplinary approach.

We’ll examine collaboration between data scientists and users within organizations, data scientists’ organizational roles, and how corporate norms influence data science.

The session will feature discussions on how organizational and individual dynamics aid data scientists in fostering beneficial collaborations and translating data insights effectively. This interactive session will include talks from a consultant, an AI entrepreneur, and a researcher, aiming to spark dialogue with attendees.

The goal is to explore current issues in corporate data science deployment and offer new perspectives and tools for participants to reflect on their collaboration with corporate data users. We’ll also discuss the evolving role of data scientists in organizations, their influence on knowledge translation, and their place in the organizational hierarchy of knowledge.


  • Introduction – Carl Stefan Roth-Kirkegaard, Postdoc, SDU 
  • Data Science… after the arrival of the large language models | Thomas Aagren Jensen, Senior Manager, PwC data science group
  • Experiences from AI startups: the role of data scientists in business – Christian Lawaetz Halvorsen, IT & AI Entrepreneur
  • Research perspective on the interplay of data scientists and business logics | Alf Rehn, Professor, SDU 
  • Comments and Q&A

Target audience 

The session is directed towards a mixed audience, i.e. both data scientists, representatives from organizations, as well as academics researching the role of data science in organizations. It is especially relevant to data scientists and organizations that are not engaged in traditional data science areas, but are working more broadly in industry or public sector which requires extensive collaboration with non-expert data users.


  • Professor Alf Rehn, SDU
  • Postdoc Carl Stefan Roth-Kirkegaard, SDU


ROOM: Salon 28+29

This session delves into the state of the art in Danish Natural Language Processing (NLP). The session will begin with an introduction to generative AI and Large Language Models (LLMs) in general, followed by an in-depth presentation of the collaborative efforts between the Alexandra Institute and the Center for Humanities Computing at Aarhus University on building a Danish-first LLM.

This first session will also provide a glimpse into future work in this domain. Next, a presentation about open problems in Danish NLP. The entire session is supposed to be interactive, with dedicated question and answer segments after both talks, and a live, hands-on session with a Danish LLM.

Main activities

  • Introduction to large language models | Alexandra Institute | 20 min.
    • The transition from expert models to general-purpose model
    • Discussion of Danish LLM use cases
  • The Danish LLM scene | CHC | 15 min.
    • Overview of projects relating to Danish LLMs
    • Model performance evaluation
  • Developing Danish LLMs: Danish Foundation Models | CHC | 15 min.
  • Hands-on workshop with Danish LLMs | CHC + Alexandra Institute | 30 min.
  • Future directions, challenges and opportunities | Alexandra | 10 min.

Target audience

The session is open to everyone interested in Danish NLP and generative AI.

We expect the session to be interesting to everyone interested in the development of Danish generative language models, as well as people interested in the potential for using these models in research or industry.

Audience cap: 50 participants.


  • Rasmus Larsen, The Alexandra Institute
  • Dan S. Nielsen, The Alexandra Institute


Both society and industry have a substantial interest in well-functioning mobility infrastructures that are efficient, predictable, environmentally friendly, and safe. For outdoor mobility, e.g., the reduction of congestion is high on the political agenda, as is the reduction of CO2 emissions, as the transportation sector is the second largest in terms of greenhouse gas emissions.

The amount of mobility-related data has increased massively, which enables an increasingly wide range of analyses. When combined with digital representations of road networks, this data holds the potential for enabling a more fine-grained understanding of mobility and for enabling more efficient, predictable, and environmentally friendly mobility.

This session will explore recent work and identify paths for future Danish collaboration across digitalization, data science, and AI within the area. This session will present recent work and discuss its future potential.

The session will discuss the topic from an interdisciplinary perspective, covering data science, big data systems, cyber[1]security, and robotics. The session will feature mainly researchers who recently joined universities in Denmark and junior researchers.

Resilient mobility is an important topic from a societal perspective, and the D3A community can make many important contributions to this agenda.

Tentative program

60-minute scientific talks, e.g.,

  1. “Towards Vision Zero: A Zero Touch Framework for Road Safety,” Professor Sadok Ben Yahia, SDU
  2. “Cyber-Security challenges for resilient Mobility, Assistant Professor Gaurav Choudhary, SDU
  3. “Data-Driven Path Planning for Mobile Robots,” PhD Student Avgi Kollakidou, SDU
  4. “Advancing Mobility Data Science,” Professor Kristian Torp, AAU
  5. “Data-driven, Distributed decision making for safer mobility,” Associate Professor Serkan Ayvaz
  6. 30 minute panel on “Future Challenges” with presenter

Target audience and size

We want to attract researchers within data science, digitalization, and robotics. We aim for 25-30 participants who will discuss the challenges and opportunities. 


  • Sadok Bin Yahia, Professor in Big Data Systems, Mærsk Mc-Kinney Møller Institute, SDU

  • Gaurav Choudhary, Assistant Professor in Cyber-Security, Mærsk Mc-Kinney Møller Institute, SDU

  • Kristian Torp, Professor in Data Management, Department of Computer Science, SDU

  • Avgi Kollakidou, PhD Student in Robotics, Mærsk Mc-Kinney Møller Institute, SDU

ROOM: Sal 22+23

This session aims to bring together machine learners and data scientists in Denmark with an interest in methodological, basic research in machine learning. Basic research is the foundation upon which science is built. Yet, machine learning and data science is naturally drawn towards applications through their reliance on data. Most available funding hinges on applications, which means most research is inherently application driven, often leaving time only to discover, but not solve, fundamental machine learning problems.

When research is tied to a downstream application, with a collaborator waiting for a solution, we often end up making specific, heuristic solutions that solve encountered problems within the particular special case, rather than solve the real underlying problems. As a result, solutions are less general, and the same problems need to be solved repeatedly rather than once and for all.

With this parallel session, we wish to bring attention to the importance of basic machine learning research, and to strengthen the community of basic machine learning researchers across Denmark.


  • Welcome | 5 mins. 
  • Talk by Oswin Krause, KU | 20 mins. 
  • Talk by Viktoria Schuster, KU | 20 mins. 
  • Talk by Kenneth Borup, AU | 20 mins. 
  • Talk by Melih Dandemir, SDU | 20 mins. 
  • Wrap up | 5 mins. 

Target audience

The target audience is Danish machine learning researchers with an interest in basic machine learning research. We would expect around 50 participants. 


The center for Basic Machine Learning Research in Life Science (, represented by: 

  • Wouter Krogh Boomsma, KU
  • Søren Hauberg, DTU
  • Jes Frellsen, DTU
  • Anders Krogh, KU
  • Aasa Feragen, DTU
  • Ole Winther, DTU and KU


Recent hardware and ML advances have enabled deep learning to harness geometry and symmetries in physics, chemistry, and biology. Graph-based learning has been used to achieve significant performance breakthroughs in particle physics, discrete symmetries are used in many state-of-the-art models in molecular chemistry, and equivariance has been used to enhance the transformer architectures of protein folding. These techniques have led to groundbreaking improvements across domains where data is structured as point clouds, graphs, or sequences. The session gathers practitioners and newcomers in these fields to share knowledge and challenges, particularly within Danish research groups. Given the technology’s varying levels of adoption among these groups, the focus is on building networks around common challenges and understanding current tools. It serves as Part 2 of the D3A ML in Science sessions.

Session Format and Finalized Program

  1. Seeds: Live demonstrations & hands-on examples of cutting-edge geometric deep learning | 45 mins.
    In this section, a set of three concrete examples will be solicited from leading researchers developing or applying Geometric Deep Learning (GDL) in the Danish physical and life sciences. These will be presented in live coding sessions, with as many of the details shown, as well as presenting any potential to apply to diverse datasets. The goal is to demystify the machinery behind geometric deep learning: graph neural networks, transformers, symmetry-constrained learning. By presenting these tools in an accessible way, with open-source libraries behind each demo, and further material for attendees to follow after the session, we hope to spur researchers new to GDL to apply these tools, and to encourage expert researchers to see how the tools are applied in other domains.
  2. Soil: Challenges in the Big Data scientific domains that could be tackled as a community | 45 mins.
    This section builds on the lightning talks given in Part 1: ML in Physical Systems. Niels Bohr Institute physicist and AI/ML researcher Troels Petersen will facilitate the session, to help connect the dots between talks, and those in the room; to initiate discussions amongst attendees; and to suggest directions for research and collaboration. We emphasize especially that we can take advantage of this conference being a ‘for-Denmark’ meeting, to encourage a high degree of openness about challenges, and willingness to learn.

Target Audience

The target audience is expected to be drawn from Danish academia and industry researching the physical and life sciences, with a focus on “young and young-at-heart” researchers (students, postdocs or staff) that are willing to cross traditional domain boundaries.

Confirmed Speakers/Presenters

  • Troels Petersen, Niels Bohr Institute


  • Daniel Murnane, Lawrence Berkeley National Laboratory

ROOM: Auditorium

The responsible implementation of ML and AI is crucial as they enter critical societal infrastructure. From healthcare, welfare benefit distribution, to selection processes such as hiring, dating, and credit scoring, AI decision support is already impacting our individual opportunities and lives. We will showcase current fairness, causality, and explainability research and discuss how they can be used to do good.

Algorithmic fairness aims to characterize, detect, diagnose, and mitigate unwanted (demographic) biases in the predictions generated by algorithms that have been fit to data. This is controversial because different notions of algorithmic fairness come with different moral prioritizations, and resolving conflicting notions is not solely a technical question. For instance, if an online job portal highlights managerial job postings to male, but not to female candidates, different fairness criteria might lead to different conclusions on whether the underlying algorithm is unfair and how it should be adapted.

Causality is important to ML whenever prediction algorithms are used to inform decisions on how to act. When ML is used to predict undesired conditions, such as disease progression, the natural desire is to mitigate such conditions. Having ML discover and fit causal relations helps such mitigation attempts be more reliable. We will discuss the potential and limitations of causality. For example, a ML model may predict high chances of developing lung cancer given a patient’s health records and lower chances if the health records were altered to show no yellow finger colourings. Based on this, we may treat and remove yellow finger colourings (removing the side-effect) and not stop smoking (removing the cause).

When using more complex data and models for less-well understood tasks, it becomes increasingly difficult to spot if a model-based recommendation is similarly misled. To what extent causality can move us toward causal ML models that better discern association and causation is controversial, not least since real-world validation is often impossible and the required assumptions are often finicky and implausible.

Explainable AI (XAI) aims to provide transparency into or explanations of models and their predictions, to help understand, inspect, and scrutinize models and certify their safety before deployment. Whether current XAI lives up to the promises of safety and accountability is debated: Countless examples have demonstrated that XAI does not live up to the expectations held by many practitioners.

We will showcase recent work from the Danish data science and adjacent communities, on aspects of algorithmic fairness, causality, and explainable AI, with a focus on young faculty, postdocs, and PhD students, both to showcase their work and to tie them into the community.

Speakers will include affiliates of the causality and explainability collaboratory from the Pioneer Center for AI, hence opening up for increased contact between the collaboratory and the broader Danish data science community. 

Responsible AI is relevant whenever AI is used to make decisions that affect humans, which covers a substantial proportion of AI applications. Causality and explainability have interest beyond the human aspect: Aiming to give insight into how variables are connected and AI predictions are made, they have great potential for aiding scientific discovery in chemistry, physics, astronomy, or climate science.

Target audience

We expect the session to cater to people developing or using XAI, causality, or algorithmic fairness, as well as people who are concerned about the safety of AI. We would like to cap the session at 50 participants to enable discussion.


  • Aasa Feragen, DTU and P1
  • Siavash Arjomand Bigdeli, DTU
  • Jens Ulrik Hansen, RUC
  • Ira Assent, AU and P1
  • Sebastian Weichwald, KU and P1


13:15-13:20 Welcome

13:20-13:40  Tareen Dawood, Kings College London:
Uncertainty & Trust in Cardiac Deep Learning Models

13:40-14:00  Mohammad Naser Sabet Jahromi, Aalborg University:
Demystifying the Black Box: The Similarity Difference and Uniqueness Method for Explaining CNN Models in Sensitive Domains

14:00-14:40  Luigi Gresele, University of Copenhagen:
Assessing Causal Reasoning in Language Models, and Why it Matters

14:35-14:55  Rolf Hvidfeldt, Aalborg University:

Fairness Unveiled: Ignorance and Responsibility in Data Science

14:40-14:45  Closing comments

ROOM: Salon 31

The climate emergency is a multidimensional and wicked problem that requires a range of responses at various scales (local, regional, national, and planetary). Whether the issue of concern is renewable energy and carbon reduction, biodiversity loss, transport, agriculture, the circular economy, mitigation and adaptation, or more desirable economies, IT plays a central role. This seminar engages with the intersection between climate and digital technologies from different disciplinary perspectives to highlight how this intersection is multilayered and fraught with tension.

Our first speaker will show how data-driven analysis can help urban planners design better infrastructure for active mobility like cycling. By addressing active mobility infrastructure from a system-wide and strategic angle, data science can identify shortcomings in, for example, cycling accessibility and help build more sustainable cities. In this talk they will present how network science can guide investments in cycling infrastructure, introduce the importance of incorporating subjective perceptions when quantifying bikeability, and finally discuss the limitations and criticism of bicycle network research.

Our second speaker will add to this perspective by looking at applications of spatiotemporal data analysis and forecasting in topics related to sustainability. Connecting to the previous talk, the speaker will give examples of data gaps in last mile logistics and how prediction models can have a positive impact on the efficiency and sustainability of delivery systems. Secondly, she will discuss spatiotemporal data processing in the use case of renewable energy, particularly wind power prediction. The speaker will discuss new machine learning approaches that leverage location information to predict wind power output and the associated costs of running more complex large models.

Finally, the third speaker will present a framing of the overall session by asking ‘what is meant by “Climate IT”?’ And discuss this concept in relation to related discussions of ’twin transition’. The paper will consequently turn to what we legitimately can expect and demand of the role of IT in green transitions and processes of sustainability. After these three short talks, there will be room for discussion with input from the audience.

Overall, the panel aims to explore the tense relations between digital technologies and climate change. On the one hand, it highlights the ways in which these technologies can create the fact-based, large-scale knowledge that can equip us to tackle the very complex problems our societies are faced with. On the other hand, the panel also critically engages with how the revolutionary promises made by digital technologies often make us gloss over their growing climate impact. Together, the different perspectives enable us to gain a deeper and more nuanced understanding of how to address arguably the most pressing challenge of our times, the climate emergency.


Seminar with three short talks for questions and discussion. 


The role of digital technologies in addressing societal challenges

Target Audience

The session is meant for anyone with an interest in climate change, both academics from different disciplaines and IT professionals. 

There is no maximun number of participants for the session. 


  • Ane Rahbeck Vierø, Ph.D. Fellow, Department of Computer Science, IT University of Copenhagen

  • Maria Sinziiana Astefanoai, Assistant Professor, Department of Computer Science, IT University of Copenhagen

  • Steffen Dalsgaard, Professor at the Department of Business IT, Head of Center for Climate IT, IT University of Copenhagen



The availability of generative AI methods and tools create both new opportunities and challenges within the IT educations. The educations need to reconsider the teaching of how to specify, design, construct, test, operate and maintain IT systems. Here Generative AI at the same time enables new means for generating the media content of interactive systems and the algorithms of data-heavy systems. Many educations also teach AI and here Generative AI systems puts even more focus on strengthen teaching in topics, such as, bias, ethics and the role of AI in the society. If this wasn’t enough Generative AI also beyond the core discipline enable new tools for supporting learning that also reframe teaching activities and university educations. This session will present recent results and discuss the challenges and experiences of integrating Generative AI in IT teaching activities and curriculums. We will combine a focus on experience reports on embracing Generative AI in individual courses with a focus on the whole educations. In a panel format the session will discuss the road ahead for generative AI and the IT educations.


Welcome and a poll on experience by Mikkel Baun Kjærgaard, Professor in Software Engineering, Mærsk Mc-Kinney Møller Institute, SDU

Generative AI in a teacher perspective:

  • Andreas Møgelmose, Assistant Professor in Visual Analytics and Perception, AAU Create: Programming and image analysis with zero experience using ChatGPT
  • Siavash Arjomand Bigdeli, Associate Professor in Computer Science, DTU Compute: Novelty in Generative AI
  • Sine Zambach: Experiences with teaching introduction to programming for business students, high school, and in continuing education – with and without AI

Generative AI in an educational perspective by Thore Husfeldt, Professor in Computer Science, Department of Computer Science, ITU

Panel debate with the speakers on “The impact of Generative AI on IT educations”

Target audience

We want to attract teachers within the D3A community. We aim for 25-30 participants that will discuss the challenges and opportunities.


  • Mikkel Baun Kjaergaard, SDU
  • Siavash Arjomand Bigdeli, CSEM
  • Andreas Møgelmose, AAU
  • Johannes Bjerva, AAU
  • Thore Husfeldt, IT
  • Sine Zambach, CBS


Human-facing Natural Language Processing

Despite the recent success of Natural Language Processing (NLP), driven by advances in large language models (LLMs) trained on enormous amounts of data, there are many challenges ahead to make human-facing NLP a reality—that is, more trustworthy and inclusive technology.

In this talk, Barbara Plank will survey some of these challenges, all related to the fascinating diversity of language. While language varies along many dimensions, the conventional machine learning paradigm relies on learning from single ground truths, disregarding the nuanced subjectivity that is an integral part of human language variation.

Another challenge relates to bridging the digital language divide and devising technology that is more inclusive and covers more language variants, particularly small languages and dialects.

In this talk, Plank will survey these challenges and outline some potential solutions.

About Barbara Plank

Professor (Chair) for AI and Computational Linguistics at LMU Munich, Co-director of the  Center for Information and Language Processing (CIS), Head of the Munich AI and Natural Language Processing (MaiNLP) research lab. She is also professor (part-time) at ITU (IT University of Copenhagen), NLP North lab. 


Closing remarks by Aasa Feragan, Technical University of Denmark