The Content-Centered Computing group at the University of Turin, Italy, offers two 14-month postdoc positions in the context of HARMONIA (Harmony in Hybrid Decision-Making: Knowledge-enhanced Perspective-taking LLMs for Inclusive Decision-Making), funded by the European Union under the NextGenerationEU program within the larger project FAIR (Future Artificial Intelligence) Spoke 2 “Integrative AI”. The project aims at developing methods for the adoption of knowledge-enhanced Large Language Models (LLMs) in supporting informed and inclusive political decisions within public decision-making processes. The PI is Prof. Viviana Patti (University of Turin).
The topics of the postdoc fellowships are:
- Computational linguistics methods for knowledge-enhanced perspective-taking LLMs to support Inclusive Decision-Making
- Perspective-taking LLMs for supporting Inclusive Decision-Making (full descriptions below)
The team includes members of the Computer Science Department and Economics and Statistics Department of the University of Turin. A PhD in Computer Science, Computational Linguistics, or related areas is highly recommended. Knowledge of Italian is not mandatory. The deadline for application is May 13th 2024.
The gross salary is €25.328 (about €1,860/month net salary). Turin is a vibrant and liveable city in Northern Italy, close to the beautiful Italian Alps and with a manageable cost of living.
Link to the call (in Italian, llok for codes INF.2024.06 and INF.2024.07). Link to the application platform. Please write to Valerio Basile or Viviana Patti for further information on how to apply.
–
Computational linguistics methods for knowledge-enhanced perspective-taking LLMs to support Inclusive Decision-Making
The activity will focus on a) design of a semantic model to represent sinteractions between urban services and citizens and integrate multi-source hybrid data; b) data annotation by citizens with different socio-cultural backgrounds to collect different perspectives on social issues. Data will be collected and organized in a Knowledge Graph. The activity will be supported by an interdisciplinary team of experts in KR, behavioral economics and LLMs (link with the design of knowledge-enhanced LLMs).
Perspective-taking LLMs for supporting Inclusive Decision-Making
The activity will focus on a) exploring techniques for integrating multi-source hybrid citizen data into LLMs (RAG and Knowledge Injection); b) developing methods for training and evaluating perspective-taking LLMs, which explicitly encode multiple perspectives, embodying the point of view of different citizen communities on a topic. Planned activities include: benchmark creation, error analysis, and evaluation of the efficiency and reliability of the developed technologies.