SciNoBo: An AI system collaborating with Journalists in Science Communication (resubmission)

Science communication conveys scientific findings and informs about research developments the general public, policymakers and other non-expert groups raising interest, trust in science and engagement on societal problems (e.g., United Nations Sustainable Development Goals). In this context, evidence-based science communication isolates topics of interest from the scientific literature, frames the relevant evidence and disseminates the relevant information to targeted non-scholarly audiences through a wide range of communication channels and strategies.

The proposed microproject (MP) focusses on science journalism and the public outreach on scientific topics in Health and Climate Change. The MP will bring together and enable interactions of science communicators (e.g., science journalists, policy analysts, science advisors for policymakers, other actors) with an AI system, capable of identifying statements about Health and Climate in mass media, grounding them on scientific evidence and simplifying the language of the scientific discourse by reducing the complexity of the text while keeping the meaning and the information the same.

Technologically, we plan to build on our previous MP work on neuro-symbolic Q&A (*) and further exploit and advance recent developments in instruction fine-tuning of large language models, retrieval augmentation and natural language understanding – specifically the NLP areas of argumentation mining, claim verification and text (ie, lexical and syntactic) simplification.

The proposed MP addresses the topic of “Collaborative AI” by developing an AI system equipped with innovative NLP tools that can collaborate with humans (ie, science communicators -SCs) communicating statements on Health & Climate Change topics, grounding them on scientific evidence (Interactive grounding) and providing explanations in simplified language, thus, facilitating SCs in science communication. The innovative AI solution will be tested on a real-world scenario in collaboration with OpenAIRE by employing OpenAIRE research graph (ORG) services in Open Science publications.

Workplan
The proposed work is divided into two phases running in parallel. The main focus in phase I is the construction of the data collections and the adaptations and improvements needed in PDF processing tools. Phase II deals with the development of the two subsystems: claim analysis and text simplification as well as their evaluation.

Phase I
Two collections with News and scientific publications will be compiled in the areas of Health and Climate. The News collection will be built based on an existing dataset with News stories and ARC automated classification system in the areas of interest. The second collection with publications will be provided by OpenAIRE ORG service and further processed, managed and properly indexed by ARC SciNoBo toolkit. A small-scale annotation is foreseen by DFKI in support of the simplification subsystem.

Phase II
In phase II, we will be developing/advancing, finetuning and evaluating the two subsystems. Concretely, the “claim analysis” subsystem encompasses (i) ARC previous work on “claim identification”, (ii) a retrieval engine fetching relevant scientific publications (based on our previous miniProject), and (iii) an evidence-synthesis module indicating whether the publications fetched and the scientists’ claims therein, support or refute the News claim under examination.
DFKI will be examining both lexical and syntax-based representations, exploring their contribution to text simplification and evaluating (neural) simplification models on the Eval dataset. Phase II work will be led by ARC in collaboration with DFKI and OpenAIRE.

Ethics: AI is used but without raising ethical concerns related to human rights and values.

(*): Combining symbolic and sub-symbolic approaches – Improving neural QA-Systems through Document Analysis for enhanced accuracy and efficiency in Human-AI interaction.

Output

Paper(s) in Conferences:
We plan to submit at least two papers about the “claim analysis” and the “text simplification” subsystems.

Practical demonstrations, tools:
A full-fledged demonstrator showing the functionality supported will be available (expected at the last month of the project).

Project Partners

  • ILSP/ATHENA RC, Haris Papageorgiou
  • German Research Centre for Artificial Intelligence (DFKI), Julián Moreno Schneider
  • OpenAIRE, Natalia Manola

Primary Contact

Haris Papageorgiou, ILSP/ATHENA RC

This project aims to make modern cognitive user models and collaborative AI tools more applicable by developing generalizable amortization techniques for them.

In human-AI collaboration, one of the key difficulties is establishing a common ground for the interaction, especially in terms of goals and beliefs. In practice, the AI might not have access to this necessary information directly and must infer it during the interaction with the human. However, training a model to support this kind of inference would require massive collections of interaction data and is not feasible in most applications.
Modern cognitive models, on the other hand, can equip AI tools with the necessary prior knowledge to readily support inference, and hence, to quickly establish a common ground for collaboration with humans. However, utilizing these models in realistic applications is currently impractical due to their computational complexity and non-differentiable structure.
This micro-project contributes directly to the development of collaborative AI by making cognitive models practical and computationally feasible to use thus enabling efficient online grounding during interaction. The project approaches this problem by developing amortization techniques for modern cognitive models and for merging them in collaborative AI systems.

Output

A conference paper draft that introduces the problem, a method, and initial findings.

Project Partners

  • Delft University of Technology, Frans Oliehoek

Primary Contact

Samuel Kaski, Delft University of Technology

SciNoBo: An AI system collaborating with Journalists in Science Communication (resubmission)

Science communication conveys scientific findings and informs about research developments the general public, policymakers and other non-expert groups raising interest, trust in science and engagement on societal problems (e.g., United Nations Sustainable Development Goals). In this context, evidence-based science communication isolates topics of interest from the scientific literature, frames the relevant evidence and disseminates the relevant information to targeted non-scholarly audiences through a wide range of communication channels and strategies.

The proposed microproject (MP) focusses on science journalism and the public outreach on scientific topics in Health and Climate Change. The MP will bring together and enable interactions of science communicators (e.g., science journalists, policy analysts, science advisors for policymakers, other actors) with an AI system, capable of identifying statements about Health and Climate in mass media, grounding them on scientific evidence and simplifying the language of the scientific discourse by reducing the complexity of the text while keeping the meaning and the information the same.

Technologically, we plan to build on our previous MP work on neuro-symbolic Q&A (*) and further exploit and advance recent developments in instruction fine-tuning of large language models, retrieval augmentation and natural language understanding – specifically the NLP areas of argumentation mining, claim verification and text (ie, lexical and syntactic) simplification.

The proposed MP addresses the topic of “Collaborative AI” by developing an AI system equipped with innovative NLP tools that can collaborate with humans (ie, science communicators -SCs) communicating statements on Health & Climate Change topics, grounding them on scientific evidence (Interactive grounding) and providing explanations in simplified language, thus, facilitating SCs in science communication. The innovative AI solution will be tested on a real-world scenario in collaboration with OpenAIRE by employing OpenAIRE research graph (ORG) services in Open Science publications.

Workplan
The proposed work is divided into two phases running in parallel. The main focus in phase I is the construction of the data collections and the adaptations and improvements needed in PDF processing tools. Phase II deals with the development of the two subsystems: claim analysis and text simplification as well as their evaluation.

Phase I
Two collections with News and scientific publications will be compiled in the areas of Health and Climate. The News collection will be built based on an existing dataset with News stories and ARC automated classification system in the areas of interest. The second collection with publications will be provided by OpenAIRE ORG service and further processed, managed and properly indexed by ARC SciNoBo toolkit. A small-scale annotation is foreseen by DFKI in support of the simplification subsystem.

Phase II
In phase II, we will be developing/advancing, finetuning and evaluating the two subsystems. Concretely, the “claim analysis” subsystem encompasses (i) ARC previous work on “claim identification”, (ii) a retrieval engine fetching relevant scientific publications (based on our previous miniProject), and (iii) an evidence-synthesis module indicating whether the publications fetched and the scientists’ claims therein, support or refute the News claim under examination.
DFKI will be examining both lexical and syntax-based representations, exploring their contribution to text simplification and evaluating (neural) simplification models on the Eval dataset. Phase II work will be led by ARC in collaboration with DFKI and OpenAIRE.

Ethics: AI is used but without raising ethical concerns related to human rights and values.

(*): Combining symbolic and sub-symbolic approaches – Improving neural QA-Systems through Document Analysis for enhanced accuracy and efficiency in Human-AI interaction.

Output

Paper(s) in Conferences:
We plan to submit at least two papers about the “claim analysis” and the “text simplification” subsystems.

Practical demonstrations, tools:
A full-fledged demonstrator showing the functionality supported will be available (expected at the last month of the project).

Project Partners

  • ILSP/ATHENA RC, Haris Papageorgiou
  • German Research Centre for Artificial Intelligence (DFKI), Julián Moreno Schneider
  • OpenAIRE, Natalia Manola

Primary Contact

Haris Papageorgiou, ILSP/ATHENA RC

Robustness verification for Concept Drift Detection

Real world data streams are rarely stationary, but subjective to concept drift, i.e., the change of distribution of the observations. Concept drift needs to be constantly monitored, so that when the trained model is no longer adequate, a new model can be trained that fits the most recent concept. Current methods of detecting concept drift typically include monitoring the performance, and triggering a signal once this drops by a certain margin. The disadvantage of this is that this method acts retroactively, i.e., when the performance has already dropped.

The field of neural network verification detects whether a neural network is susceptible to an adversarial attack, i.e., whether a given input image can be perturbed by a given epsilon, such that the output of the network changes. This indicates that this input is close to the decision boundary. When the distribution of images that are close to the decision boundary significantly changes, this indicates that concept drift is occurring, and we can proactively (before the performance drops) retrain the model. The short-term goal of this micro-project is to define ways to a) monitor the distribution of images close to the decision boundary, and b) define control systems that can act upon this notion.

Disadvantages of this are that verifying neural networks requires significant computation time, and it will take many speed-ups before this can be utilized in high-throughput streams.

Output

Conference or Journal Paper – We initially aim for top-tier venues, but will decide on the actual venue after the results and scope are determined.

Project Partners

  • Leiden University, Holger Hoos
  • INESC TEC, João Gama

Primary Contact

Holger Hoos, Leiden University

Extending Inverse Reinforcement Learning to elicit and exploit richer expert feedback by leveraging the learner’s beliefs.

Interactive Machine Learning (IML) has gained significant attention in recent years as a means for intelligent agents to learn from human feedback, demonstration, or instruction. However, many existing IML solutions primarily rely on sparse feedback, placing an unreasonable burden on the expert involved. This project aims to address this limitation by enabling the learner to leverage richer feedback from the expert, thereby accelerating the learning process. Additionally, we seek to incorporate a model of the expert to select more informative queries, further reducing the burden placed on the expert.

Objectives:
(1) Explore and develop methods for incorporating causal and contrastive feedback, as supported by evidence from psychology literature, into the learning process of IML.
(2) Design and implement a belief-based system that allows the learner to explicitly maintain beliefs about the possible expert objectives, influencing the selection of queries.
(3) Utilize the received feedback to generate a posterior that informs subsequent queries and enhances the learning process within the framework of Inverse Reinforcement Learning (IRL).

The project addresses several key aspects highlighted in the workpackage on Collaboration with AI Systems (W1-2). Firstly, it focuses on AI systems that can communicate and understand descriptions of situations, goals, intentions, or operational plans to establish shared understanding for collaboration. By explicitly maintaining beliefs about the expert’s objectives and integrating causal and contrastive feedback, the system aims to establish a common ground and improve collaboration.
Furthermore, the project aligns with the objective of systems that can explain their internal models by providing additional information to justify statements and answer questions. By utilizing the received feedback to generate a posterior and enhance the learning process, the system aims to provide explanations, verify facts, and answer questions, contributing to a deeper understanding and shared representation between the AI system and the human expert.
The project also demonstrates the ambition of enabling two-way interaction between AI systems and humans, constructing shared representations, and allowing for the adaptation of representations in response to information exchange. By providing tangible results, such as user-study evaluations and methods to exploit prior knowledge about the expert, the project aims to make measurable progress toward collaborative AI.

Output

(1) Identification and development of potential informative feedback mechanisms that are more user-friendly, with a focus on determining the appropriate form of queries.
(2) User-study evaluation results that measure the correctness of the information provided by the human and assess the cognitive overhead involved.
(3) Methods to exploit prior knowledge about the expert to improve learning and reduce the burden placed on them, specifically in terms of how to query.
(4) Integration of richer feedback from the expert, including causal knowledge and contrastive information, into the learning process.
(5) Publication of a peer-reviewed paper in a competitive venue, presenting the research findings and contributions to the field.
(6) Creation of a GitHub repository containing all necessary materials to replicate the results and support further research endeavors.

Project Partners

  • ISIR, Sorbonne University, Silvia Tulli
  • Colorado State University, Sarath Sreedharan
  • ISIR, Sorbonne University, Mohamed Chetouani

Primary Contact

Silvia Tulli, ISIR, Sorbonne University

Accelerating nurse training without impacting the quality of education, by leveraging LWM (large whatever models) to provide individual feedback to students and help teachers how to optimize teaching.

High-quality education and training of nurses are of utmost importance to keep high standards in medical care. Nevertheless, as the covid pandemic has shown quite impressively, there are too few healthcare professionals available. Therefore, education and training of nurse students, or adapting the training of nurses is challenged to accelerate, to have manpower of nurses available when it is required. Still, accelerating training often comes with reduced quality, which can easily lead to bad qualifications and, in the worst case, to a lethal outcome.
Thus, in nurse training a pressing question is, how to optimize and with it accelerate training without suffering in quality.
One of the significant questions for teachers in training nurse students is to understand the state of a student’s education. Are some students in need of more repetitions? which students can proceed to the next level, who is ready to get in contact with actual patients? In this regard, optimization of training means to individualize, not only individualize the training of students but also individualize the feedback and information a teacher gets about their way of teaching.

We believe this to be a field where Artificial Intelligence (AI) and more specifically the application of foundational models (LLMs large language models, paired with other methods of machine learning) can provide real support.

In the first part of this microproject, together with Nurse-Teachers of the University of Southampton, we want to define and design an LWM that fits the requirements of nurse training. For this, 2-3 nurse teachers from Southampton will visit DFKI in order to get a feeling for systems that are available, and also what applications are feasible. In turn, researchers of DFKI will visit the nurse training facilities in Southampton to get a better picture of how nurse training is conducted. At the end of this first phase of the microproject, an LWM (large whatever model) is defined (existing LLMs combined with additional features and data sources, as required).

In the second phase, this LWM will be implemented and tested against videos of recorded training sessions. Specific focus will be set on:
• How to understand the action of a particular person?
• Actions taken by the trainee, are they correct or false? What would have been the correct action?
• Which teaching efforts work and which do not as much?
• Which useful suggestions and feedback can be provided to the trainees and teachers?

Depending on the outcome of this microproject, in a follow-up project, an online LWM system could be installed at the facilities of the University of Southampton, where the effects of direct feedback on teaching and performance, could be evaluated.

Output

1) Definition and design of the LWM will be documented and if possible published in an adequate scientific journal
2) Developed algorithms and results will be published at a scientific conference (AI and possibly also medical)
3) The developed LWM will be made available to be used in a follow-up project

Project Partners

  • DFKI, EI, Agnes Grünerbl
  • Health Department, Unviversity of Southampton, Eloise Monger

Primary Contact

Agnes Grünerbl, DFKI, EI

Build human-in-the-loop intelligent systems for the geolocation of social media images in natural disasters

Social media generate large amounts of almost real-time data which can turn out extremely valuable in an emergency situation, specially for providing information within the first 72 hours after a disaster event. Despite there is abundant state-of-the-art machine learning techniques to automatically classify social media images and some work for geolocating them, the operational problem in the event of a new disaster remains unsolved.
Currently the state-of-the-art approach for dealing with these first response mapping is first filtering and then submitting the images to be geolocated to a crowd of volunteers [1], assigning the images randomly to the volunteers.

The project is aimed at leveraging the power of crowdsourcing and artificial intelligence (AI) to assist emergency responders and disaster relief organizations in building a damage map from a zone recently hit by a disaster.

Specifically, the project will involve the development of a platform that can intelligently distribute geolocation tasks to a crowd of volunteers based on their skills. The platform will use machine learning to determine the skills of the volunteers based on previous geolocation experiences.

Thus, the project will concentrate on two different tasks:
• Profile Learning. Based on the previous geolocations of a set of volunteers, learn a profile of each of the volunteers which encodes its geolocation capabilities. This profiles should be unterstood as competency maps of the volunteer, representing the capability of the volunteer to provide an accurate geolocation for an image coming from a specific geographical area.
• Active Task Assigment. Use the volunteer profiles efficiently in order to maximize the geolocation quality while maintaining a fair distribution of geolocation tasks among volunteers.

On a first stage we envision an experimental framework with realistically generated artificial data, which acts as a feasibility study. This will be published as a paper in a major conference or journal. Simultaneously we plan to integrate both the profile learning and the active task assignment with the crowdnalysis library, a software outcome of our previous micro-project. Furthermore, we plan to organize a geolocation workshop to take place in Barcelona with participation from the JRC, University of Geneva, United Nations, and IIIA-CSIC.

In the near future, the system will generate reports and visualizations to help these organizations quickly understand the distribution of damages. The resulting platform could enable more efficient and effective responses to natural disasters, potentially saving lives and reducing the impact of these events on communities.
The microproject will be developed by IIIA-CSIC and the University of Geneva. The micro project is also of interest to the team lead by Valerio Lorini at the Joint Research Center of the European Commission @ Ispra, Italy, who will most likely attend the geolocation workshop which we will be putting forward.

The project is in line with “Establishing Common Ground for Collaboration with AI Systems (WP 1-2)”, because it is a microproject that ” that seeks to provide practical demonstrations, tools, or new theoretical models for AI systems that can collaborate with and empower individuals or groups of people to attain shared goals” as is specifically mentioned in the Call for Microprojects.

The project is also in line with “Measuring, modeling, predicting the individual and collective effects of different forms of AI influence in socio-technical systems at scale (WP4)” since it ecomprises the design of a human-centered AI architectures that balance individual and collective goals for the task of geolocation.

[1] Fathi, Ramian, Dennis Thom, Steffen Koch, Thomas Ertl, and Frank Fiedrich. “VOST: A Case Study in Voluntary Digital Participation for Collaborative Emergency Management.” Information Processing & Management 57, no. 4 (July 1, 2020): 102174. https://doi.org/10.1016/j.ipm.2019.102174.

Output

– Open source implementation of the volunteer profiling and consensus geolocation algorithms into the crowdnalysis library.
– Paper with the evaluation of the different geolocation consensus and active strategies for geolocation
– Organization of a one day workshop with United Nations, JRC, University of Geneva, CSIC

Project Partners

  • Consejo Superior de Investigaciones Científicas (CSIC), Jesus Cerquides
  • University of Geneva, Jose Luis Fernandez Marquez

Primary Contact

Jesus Cerquides, Consejo Superior de Investigaciones Científicas (CSIC)

Develop AI interactive grounding capabilities in collaborative tasks using a game-based mixed reality scenario that require physical actions.

The project addresses research on interactive grounding. It consists of the development of an Augmented Reality (AR) game, using HoloLens, that supports the interaction of a human player with an AI character in a mixed reality setting using gestures as the main communicative act. The game will integrate technology to perceive human gestures and poses. The game will bring about collaborative tasks that need coordination at the level of mutual understanding of the several elements of the required task. Players (human and AI) will have different information about the tasks to advance in the game and need to communicate that information to their partners through gestures. The main grounding challenge will be based on learning the mapping between gestures to the meaning of actions to perform in the game. There will be two levels of gestures to ground, some are task-independent while others are task-dependent. In other words, besides the gestures that communicate explicit information about the game task, the players need to agree on the gestures used to coordinate the communication itself, for example, to signal agreement or doubt, to ask for more information, or close the communication. These latter gesture types can be transferred from task to task within the game, and probably to other contexts as well.
It will be possible to play the game with two humans and study their gesture communication in order to gather the gestures that emerge: a human-inspired gesture set will be collected and serve the creation of a gesture dictionary in the AI repertoire.
The game will provide different tasks of increasing difficulty. The first ones will ask the players to perform gestures or poses as mechanisms to open a door to progress to the next level. But later, in a more advanced version of the game, specific and constrained body poses, interaction with objects, and the need to communicate more abstract concepts (e.g., next to, under, to the right, the biggest one, …) will be introduced.
The game will be built as a platform to perform studies. It will support studying diverse questions about the interactive grounding of gestures. For example, we can study the way people adapt to and ascribe meaning to the gestures performed by the AI agent, we can study how different gesture profiles influence the people’s interpretation, facilitate grounding, and have an impact on the performance of the tasks, or we can study different mechanisms on the AI to learn its gesture repertoire from humans (e.g., by imitation grounded on the context).
We see this project as a relevant contribution to the upcoming Macro Project on Interactive Grounding, and we would like the opportunity to join the MP later. Our focus is on the grounding based on gestures being critical in certain scenarios. The setting can include language if vocalization is allowed and can be heard. Our game scenarios are simple and abstract and can be the basis for realistic ones.

Output

A game that serves as a platform for studying grounding in the context of collaborative tasks using gestures.
A repertoire of gestures to be used in the communication between humans and AI in a collaborative task that relies on the execution of physical actions. We will emphasize the gestures that can be task-independent.
The basis for an AI algorithm to ground gestures to meaning adapted to a particular user.
One or two papers, describing the platform and a study with people.

Project Partners

  • Instituto Superior Técnico (IST), Rui Prada
  • Eötvös Loránd University, András Lőrincz
  • DFKI Lower Saxony, Daniel Sonntag
  • CMU, László Jeni

Primary Contact

Rui Prada, Instituto Superior Técnico (IST)

A graduate level educational module (12 lectures + 5 assignments) covering basic principles and techniques of Human-Interactive Robot Learning.

Human-Interactive Robot Learning (HIRL) is an area of robotics that focuses on developing robots that can learn from and interact with humans. This educational module aims to cover the basic principles and techniques of Human-Interactive Robot Learning. This interdisciplinary module will encourage graduate students (Master/PhD level) to connect different bodies of knowledge within the broad field of Artificial Intelligence, with insights from Robotics, Machine Learning, Human Modelling, and Design and Ethics. The module is meant for Master’s and PhD students in STEM, such as Computer Science, Artificial Intelligence, and Cognitive Science.
This work will extend the tutorial presented in the context of the International Conference on Algorithms, Computing, and Artificial Intelligence (ACAI 2021) and will be shared with the Artificial Intelligence Doctoral Academy (AIDA). Moreover, the proposed lectures and assignments will be used as teaching material at Sorbonne University, and Vrije Universiteit Amsterdam.
We plan to design a collection of approximately 12 1.5-hour lectures, 5 assignments, and a list of recommended readings, organized along relevant topics surrounding HIRL. Each lecture will include an algorithmic part and a practical example of how to integrate such an algorithm into an interactive system.
The assignments will encompass the replication of existing algorithms with the possibility for the student to develop their own alternative solutions.

Proposed module contents (each lecture approx. 1.5 hour):
(1) Interactive Machine Learning vs Machine Learning – 1 lecture
(2) Interactive Machine Learning vs Interactive Robot Learning (Embodied vs non-embodied agents) – 1 lecture
(3) Fundamentals of Reinforcement Learning – 2 lectures
(4) Learning strategies: observation, demonstration, instruction, or feedback
– Imitation Learning, Learning from Demonstration – 2 lectures
– Learning from Human Feedback: evaluative, descriptive, imperative, contrastive examples – 3 lectures
(5) Evaluation metrics and benchmarks – 1 lecture
(6) Application scenarios: hands-on session – 1 lecture
(7) Design and ethical considerations – 1 lecture

Output

Learning objectives along Dublin descriptors:
(1) Knowledge and understanding;
– Be aware of the human interventions in standard machine learning and interactive machine learning.
– Understand human teaching strategies
– Gain knowledge about learning from feedback, demonstrations, and instructions.
– Explore ongoing works on how human teaching biases could be modeled.
– Discover applications of interactive robot learning.
(2) Applying knowledge and understanding;
– Implement HIRL techniques that integrate different types of human input
(3) Making judgments;
– Make informed design choices when building HIRL systems
(4) Communication skills;
– Effectively communicate about own work both verbally and in a written manner
(5) Learning skills;
– Integrate insights from theoretical material presented in the lecture and research papers showcasing state-of-the-art HIRL techniques.

Project Partners

  • ISIR, Sorbonne University, Mohamed Chetouani
  • ISIR, Sorbonne University, Silvia Tulli
  • Vrije Universiteit Amsterdam, Kim Baraka

Primary Contact

Mohamed Chetouani, ISIR, Sorbonne University

Presenting users with explanations about AI systems, to let them detect and mitigate discriminatory patterns

Our project revolves around the topic of fair Artificial Intelligence (AI), a field that explores how decision-making algorithms used in high-stake domains, such as hiring or loan allocations can perpetuate discriminatory patterns in the data they are based on, unfairly affecting people of certain races, genders or other demographics.
Early attempts to address bias in AI systems focused on automated solutions, attempting to eliminate discrimination by establishing mathematical definitions of "fairness" and optimizing algorithms accordingly. However, these approaches have faced justified criticism for disregarding the contextual nuances in which algorithms operate and for neglecting the input of domain experts who understand and can tackle discriminatory patterns effectively. Consequently, policymakers have recognized the pitfalls of solely relying on these approaches and are now designing legal regulations, mandating that high-risk AI systems can only be deployed when they allow for oversight and intervention by human experts.

With our project, we investigate how to effectively achieve this human control, by exploring the intersection between fair and explainable AI (xAI), whereas the latter is concerned with explaining the decision processes of otherwise opaque black-box algorithms. We will develop a tool, that provides humans with explanations about an algorithmic decision-making system. Based on the explanations users can give feedback about the system’s fairness and choose between different strategies to mitigate its discriminatory patterns. By immediately getting feedback about the effects of their chosen strategy, users can engage in an iterative process further refining and improving the algorithm. Since little prior work has been done on Human-AI collaboration in the context of bias mitigation, we will take on an exploratory approach to evaluate this system. We will set up a think-aloud study where potential end-users can interact with the system and try out different mitigation strategies. We will analyse their responses and thoughts, to identify the tool’s strengths and weaknesses as well as users’ mental model of the tool. Additionally, we will compare the systems’ biases before and after human intervention, to see how biases were mitigated and how successful this mitigation was.

Our work aligns with the goal of the topic “Establishing Common Ground for Collaboration with AI Systems“ (motivated by Workpackage 1 and 2). This topic is focused on developing AI systems that work in harmony with human users, empowering them to bring their expertise and domain knowledge to the table. In particular, our work recognizes humans’ ability to make ethical judgements and aims to leverage this capability to make fairer AI systems. By conducting a user study we align with the topics’ goal to make this human-AI collaboration desirable from the users’ site, ensuring that they understand the inner workings of the AI system and they have full control in adapting it.

Output

– A tool that presents users with explanations about a decision-making system and that can interactively adjust its decision process, based on human feedback about the fairness of its explanations
– A user-centric evaluation of the tool, investigating whether users can effectively detect biases through the tool and how they use the different bias mitigation strategies offered by it
– We aim to present a demo of the tool at a workshop or a conference
– Additionally, we commit to publishing one paper, describing the tool itself and the results of the usability study

Project Partners

Primary Contact

Dino Pedreschi, University of Antwerp – Departement of CS

Using Dynamic Epistemic Logic (DEL) so an AI system proactively can make announcements to avoid undesirable future states based on the human's false belief

Previously we have investigated how an AI system can be proactive, that is, acting anticipatory and on own initiative, by reasoning on current and future states, mental simulation of actions and their effects, and what is desirable. In this micro-project we want to extend our earlier work doing epistemic reasoning. That is, we want to do reasoning on knowledge and belief of the human and by that inform the AI system what kind of proactive announcement to make to the human. As in our previous work, we will consider which states are desirable and which are not, and we too will take into account how the state will evolve into the future, if the AI system does not act. Now we also want to consider the human's false beliefs. It is not necessary and, in fact, not desirable to make announcements to correct each and any false belief that the human may have. For example, if the human is watching the TV, she need not be informed that the salt is in the red container and the sugar is in the blue container, while the human's belief is that it is the other way around. On the other hand, when the human starts cooking and is about to use the content of the blue container believing it is salt, then it is a relevant announcement of the AI system to inform the human what is actually the case to avoid undesirable outcomes. The example shows, that we need to research on not only what to announce but also when to make the announcement.

The methods we will use in this micro-project are knowledge-based, to be precise, we will employ Dynamic Epistemic Logic (DEL). DEL is a modal logic. It is an extension of Epistemic Logic which allows to model change in knowledge and belief of an agent herself and of other agents.

1 week of visit is planned. In total, 7,5 PMs are planned to work on the MP, that is, 1 week we work physically in the same place, the rest of the PMs we work together online.

Output

– Formal model
We expect to develop a formal model based on DEL and based on the
findings of J.Grosinger's previous work on proactivity. The model
enables an artificial agent to make announcements to the human to
correct the human's false belief and false belief about desirability
of future states in a proactive way. Being formal we can make general
definitions and propositions in the model and provide proofs about its
properties, for example, about which proactive announcements are
relevant and/or well-timed.

– Conference
We aim for a publication of our work at an international peer-reviewed
high-quality conference. Candidate conferences are AAMAS
(International Conference on Autonomous Agents and Multiagent
Systems), or if this is temporally infeasible, then IJCAI
(International Joint Conferences on Artificial Intelligence).

– Further collaboration
The MP can lead to further fruitful collaborations between the
applicants (and possibly, some of their colleagues additionally) as
the MP's topic is new and under-explored and all cannot be investigated
within one MP.

Project Partners

  • Örebro University, ORU, Jasmin Grosinger
  • Denmark Technical Unisersity, Thomas Bolander

Primary Contact

Jasmin Grosinger, Örebro University, ORU

AI/ML methods to provide interpretable explanations and new knowledge for rare diseases.

To date, we know more than 7000 rare diseases and for the majority of them, there is a lack of relevant and quality data, also due to the fact that for a particular rare disease, there are only a few patients diagnosed in the world (small cohorts) and as these patients are living all across the globe it is difficult to perform clinical observations and upon this clinical data collection. On the other hand, due to the rapid development in gene therapies, there is also increased interest in disease-specific data from the biotech and pharma companies, but it is very hard to conduct them. However, there has been some positive shift in the last years in relation to data collection (with platforms collecting rare diseases specific data). These data are not collected in the clinical setting and are labelled as real-world data (RWD) as these data represent real insights and are conducted by citizens. RWD are not only lifestyle data (diet, sleep monitoring, etc.) collected through fitness trackers and smartwatches, but also PROs (patient/caregiver reported data). Specific rare disease platforms that collect PROs are usually using already approved/validated questionnaires. Due to the fact that patients/caregivers can answer questions online and on their own pace, these data platforms are very convenient to reach as many patients with a specific rare disease as possible (the global aspect), which is so hard to reach with classical in-person clinical settings. However, the collected data are not yet fully exploited, as platforms are mainly focusing on data collection only and not on data analytics. Because of that, the full potential of the PROs for rare diseases is still yet to come. In addition, clinicians are also not yet convinced that RWD PROs could be used for clinical research work, and this is something that we would like to change. The main objective is to develop AI/ML methods to provide interpretable explanations and new knowledge for rare diseases. The focus will be on the research of AI/ML methodologies on top of PROs, with the aim to show what information the collected data contains, and how to present this data to the clinicians in a structured, insightful, and helpful way. Our use case is the Genida registry (Genetic of Intellectual Disability and Autism Spectrum Disorders registry, managed by external partner IGBMC), collecting caregiver-reported data, as the rare disease patients covered are children and/or adults with intellectual disabilities. Our specific focus is the Kleefstra syndrome cohort, involving data for 200 Kleefstra syndrome patients from all continents. Till today this data represents the largest database of Kleefstra syndrome patients and their clinical features. Another important feature is that Genida is collecting data on a longitudinal basis, that is why correlations of symptoms during different time frames could be researched. For better UX, we will also build on human-computer interaction. This will be done in the sense of showing the results to the user (e.g. clinician), and the user would have a chance to ask the system back about the results and how and why the results were conducted. The system would show the features that help with the result explanation (e.g. which words were the most frequent in the cluster). As Kleefstra syndrome was discovered in year 2010 by clinical geneticist prof Tjitske Kleefstra from Netherlands (external partner Erasmus MC), it is relatively new. Kleefstra syndrome belongs to the group of neurodevelopmental disorders (short NDDs). With the rapidly evolving field of genetics, especially the technological advancements in genome sequencing, it is no wonder that NDDs represent the majority of rare diseases. Now it is time for AI/ML methodologies to thrive with new insights that are so much needed, as all of these diseases are so immensely underresearched.

Output

This micro project will develop new AI/ML research methodologies enabling new insights into rare diseases. The Kleefstra syndrome cohort involving data for 200 Kleefstra syndrome patients from all over the world will serve as our use case and the developed research results will be presented as a good practice example to clinicians, researchers, and rare disease patient advocacy organizations. With the results, we want to encourage further and wider participation of patients/caregivers in the data collection processes and the involvement of this data in the clinical and research work of clinicians and researchers. For better UX, we will build also on human-computer interaction ideas. This will be done in the sense of showing the micro project results to the user (e.g. clinician) using an user interface (UI), and the user would have a chance to ask the system back why the results are like that. The system would show the features that help with the result explanation (e.g. which words were the most frequent in the cluster). Main results of the micro project: The developed research methodologies will enable new insights into rare diseases through data analysis and AI/ML, and will serve the whole rare disease community. Tangible outputs:
– scientific publication
– a tangible result will be made available through the AI4EU (AI4Europe) platform

Project Partners

  • Jožef Stefan Institute, Erik Novak
  • Erasmuc MC, Tjitske Kleefstra
  • IGBMC, Pauline Burger
  • IDefine Europe, Martin Draksler

Primary Contact

Tanja Zdolšek Draksler, Jožef Stefan Institute