PhD thesis on the formalization and integration of well-being into recommendation algorithms (M/F)

Job Description

Thesis Subject

Phd position

The performance of recommendation algorithms that make use of human behavior heavily depends on their ability to capture the experience of people who interact with them. For instance, in Education, recommendations of courses and tests should consider not only the learners' evolving knowledge but also aspects of their well-being such as engagement, motivation, interest, satisfaction, frustration, boredom, fatigue and anxiety [0]. Various studies have shown that integrating well-being into decision-making generates better mental health, less burnout, and better performance over time [3,6]. The work of the candidate will be in the context of FeelGoodAI, a project with the goal of rethinking recommendation approaches to represent and leverage human well-being in decision-making.

FeelGoodAI will be applied to the field of education, where learner mental health and well-being are major concerns. Multiple perspectives and constructs related to well-being underpin several theories from positive psychology currently adopted in AI for Education (AIED). These include Flow Theory [4], characterized by a deep immersion in suitably challenging, goal-directed activities (see enclosed figure), and Self-determination Theory [25], which highlights the importance of autonomy, competence, and relatedness in learner motivation. The goal of the candidate is to to formalize well-being in education and to design AI orchestration algorithms capable of sustaining mastery progression while regulating learner well-being.

Missions
The candidate will have several objectives:

Computational formalization of Well-Being : Define a multidimensional well-being framework including frustration, anxiety, engagement, boredom, etc and model these dimensions as latent dynamical states, temporally persistent states and modulators of performance. The candidate will have to extend student models such as IRT, BKT accordingly but also estimate well-being from real Platforms by extracting implicit indicators (response time, abandonment, comments, peer interaction)

Formalization of algorithms for orchestrating educational AI agents : Train RL and LLM agents and study multi-objective optimization (mastery, well-being stability)

Work with an engineer to deploy controlled experiments and collect well-being dimensions.

Activities
- Formal modeling of well-being
- Implementation of enriched student simulator
- Development of multi-objective RL agents
- Optional LLM integration for strategy generation
- Deployment on partner platforms
- Longitudinal experimental analysis
- Participation in publications and project deliverables

Expected Contributions
- A formal framework for modeling well-being in adaptive learning
- A well-being-aware student simulation model
- A multi-objective RL framework
- An orchestration architecture
- Experimental validation on simulated and real data

Skills
Abstraction capabilities, strong programming skills in C/C++ and Python, collaboration skills. English is needed.

REFERENCES
[0] C. Bekker, S. Rothmann, M. Kloppers (2023). The happy learner: Effects of academic boredom, burnout, and engagement, Frontiers in Psychology 13, (2023)
[1] Abdin, M. et al. (2024). Phi-3: A Highly Capable Language Model Locally on Your Phone. arXiv:2404.14219
[2] Amer-Yahia S. 2024. Intelligent Agents for Data Exploration. Proc. VLDB Endow. 17, 12 (2024).
[3] Black, A. E., Deci, E. L. (2000). The effects of Instructors' Autonomy Support and learners' Autonomous
Motivation on Learning Organic Chemistry. Science Education, 84(6), 740–756.
[4] Basawapatna A., Repenning A., Koh H., Nickerson H. (2013). The Zones of Proximal Flow: Guiding students through a Space of Computational Thinking Skills. ICER 2013 67-74.
[5] Besta M. et al. (2024). Graph of Thoughts: Solving Elaborate Problems with LLMs. AAAI Conference on AI.
[6] Bittencourt, Ig. et al. (2023). Positive AI in Education (P-AIED): A Roadmap. Journal of AIED.
[7] Bonino G., Sanmartino G., Gatti-Pinheiro, Papotti P ., Troncy R., Michiardi P . (2024). Fine Tuning a Large Language Model for Socratic Interactions. Workshop on AI for Education (AI4EDU).
[8] Bouarour N., Benouaret I., Amer-Yahia S. (2024). Multi-Objective Test Recommendation for Adaptive
Learning. Trans. Large Scale Data Knowl. Centered Syst. 1-36.
[9] Chiang D., Lee H. (2023). Can LLMs be an Alternative to Human Evaluations? Assoc. for Comp. Linguistics.
[10] Durante Z., Huang Q., Wake N., Gong R., Park J. S., Sarkar B., Taori R., Noda Y ., Terzopoulos D., Choi Y ., (2024). Agent AI: Surveying the Horizons of Multimodal Interaction. arXiv:2401.03568
[11] Deeva G., Bogdanova D., Serral E., Snoeck M., De Weerdt J.( 2021). A Review of Automated Feedback Systems for Learners. Computers & Education 162.
[12] Gao J., Galley M., Li L. (2018). Neural Approaches to Conversational AI. Association for Comp. Linguistics.
[13] Guinet G., Omidvar-Tehrani B., Deoras A., Callot L. (2024). Automated Evaluation of Retrieval-Augmented LLMs with Task-Specific Exam Generation. arXiv:2405.13622
[14] Hao S., Gu Y ., Ma H., Hong J., Wang D., Hu. Z. (2023). Reasoning with LLMs is Planning with Word Model. EMNLP .
[15] Heutte, J., Fenouillet, F., Martin-Krumm, C., Gute, G., Raes, A. Gute, D., Bachelet, R. & Csikszentmihalyi, M. (2021). Optimal Experience in Adult Learning: Validation of Flow in Education. Frontiers in Psychology, 12, 1-12.
[16] Hong J., Lee N., Thorne J. (2024). ORPO: Monolithic Pref. Opt. without Reference Model. arXiv:2403.07691
[17] Huang X., Liu W., Chen X., Wang X., Wang H., Lian D., Wang Y ., Tang R., Chen E. (2024). Understanding the Planning of LLM Agents: A survey. arXiv:2402.02716
[18] Liévin V., Hother C. E., Motzfeldt A. G., Winther O. (2022). Can LLMs Reason about Medical Questions? arXiv:2207.08143
[19] Liu Y ., Iter D., Xu Y ., Wang S., Xu R., Zhu C. (2023). G-Eval: NLG Eval. GPT-4 with Human Alignment. EMNLP .
[20] Matsubara M., Borromeo R., Amer-Yahia S., Morishima A. (2021). Task assignment strategies for Crowd Worker Ability Improvement. ACM Hum. Comput. Interact. 5 (CSCW): 1-375.
[21] McKee, K.R., Tacchetti, A., Bakker, M.A. et al. Scaffolding Cooperation in Human Groups with Deep Reinforcement Learning. Nat Hum Behav 7, 1787–1796 (2023).
[22] Melnyk I., Mroueh Y ., Belgodere B., Rigotti M., Nitsure A., Yurochkin M., Greenewald K., Navratil J., Ross J. (2024). Distributional Preference Alignment of LLMs via Optimal Transport. arXiv:2406.05882
[23] Pilourdault J., Amer-Yahia S., Basu Roy S. Lee D. (2023). Task Relevance and Diversity as Worker Motivation in Crowdsourcing. IEEE ICDE.
[24] Rafailov R., Sharma A., Mitchell E., Manning C. D., Ermon S., Finn C. (2023). Direct Preference Optimization: Your Language Model is Secretly a Reward Model. NeuRIPS.
[25] Ryan, R. M., Deci, E. L. (2000). Self-Determination Theory and the Facilitation of Intrinsic Motivation, Social Development, and Well-Being. American Psychologist, 55, 68-78.
[26] Schulman J., Wolski F., Dhariwal P ., Radford A., Klimov O. (2017). PPO Algorithms. arXiv:1707.06347
[27] Shankar S., J. D. Zamfirescu-Pereira, Hartmann B., Parameswaran A., Arawjo I. (2024). Who Validates the Validators? Aligning LLM-Assisted Evaluation of LLM Outputs with Human Preferences. CoRR, abs/2404.12272.
[28] Wang C., Deng Y ., Lyu Z., Zeng L., He J., Yan S., An. B. (2024). Q*: Improving Multi-step Reasoning for LLMs with Deliberative Planning. arXiv:2406.14283
[29] Wei J., Wang X., Schuurmans D., Bosma M., Ichter B., Xia F., Chi E. H., Le Q. V., Zhou. D. (2022). Chain-of-Thought Prompting Elicits Reasoning in LLMs. NeuRIPS.
[30] Yao S., Yu D., Zhao J., Shafran I., Griffiths T. L., Cao Y ., Narasimhan. K. R. (2023). Tree of Thoughts: Deliberate Problem Solving with LLMs. NeuRIPS.
[31] Chen Y ., Wang R., Jiang H., Shi S., Xu R. Exploring the Use of LLMs for Reference-Free Quality Evaluation. (2023). CoRR, abs/2304.00723.
[32] W. H. Nori, Y . T. Lee, E. Horvitz. Can Generalist Foundation Models Outcompete Special-Purpose Tuning? Case Study in Medicine. In CoRR, abs/2311.16452, 2023.
[33] S. Lamsiyah, A. El Mahdaouy. A Reinforcement Learning-based Method for Educational Question Generation. In Artificial Intelligence in Education. AIED 2024.
[34] K. Ethayarajh, W. Xu, N. Muennighoff, D. Jurafsky, W. Kiela. KTO: Model Alignment as Prospect Theoretic Optimization. arXiv:2402.01306, 2024.

Your Work Environment

The work will take place at the Grenoble Informatics Lab (LIG), a 450-member laboratory with teaching faculty, full-time researchers, PhD students, administrative and technical staff. The mission of LIG is to contribute to the development of fundamental aspects of Computer Science (models, languages, methodologies, algorithms) and address conceptual, technological, and societal challenges. The 22 research teams in LIG aim to increase diversity and dynamism of data, services, interaction devices, and use cases influence the evolution of software and systems to guarantee the essential properties such as reliability, performance, autonomy, and adaptability. Research within LIG is organized into 5 focus areas: Intelligent Systems for Bridging Data, Knowledge and Humans, Software and Information System Engineering, Formal Methods, Models, and Languages, Interactive and Cognitive Systems, Distributed Systems, Parallel Computing, and Networks.

The host team, DAISY, is a joint CNRS, Grenoble INP, and UGA research team handling research challenges at the intersection of AI and data management, but also when data is sourced from interdisciplinary domains such as education and health.

The position is located in an area subject to French legislation on the protection of scientific and technical potential (PPST), and therefore requires, in accordance with regulations, that your arrival be authorized by the competent authority of the Ministry of Higher Education and Research (MESR).

Compensation and benefits

Compensation

2300 € gross monthly

Annual leave and RTT

44 jours

Remote Working practice and compensation

Pratique et indemnisation du TT

Transport

Prise en charge à 75% du coût et forfait mobilité durable jusqu’à 300€

About the offer

About the CNRS

The CNRS is a major player in fundamental research on a global scale. The CNRS is the only French organization active in all scientific fields. Its unique position as a multi-specialist allows it to bring together different disciplines to address the most important challenges of the contemporary world, in connection with the actors of change.

CNRS

The research professions