Vacant job

For the full announcement, click on "Apply via external website".

The Research Project

This project investigates human-AI collaboration in decision-making from the perspective of dynamic emotional adaptation. The overarching goal is to improve the outcomes of collaboration and the quality of individual human decisions.

As a researcher in this project, you will work on how theories of human affective processing can be implemented and tested in artificial agents. A central question is how systems consisting of a human and one or more artificial agents can communicate through bidirectional emotional signaling. For this to be possible, artificial agents must be able to both express and model emotions in ways meaningfully adapted to human affective processing.

Study areas may include decisions based on personal preferences, such as buying a house; decisions and assessments within legal or other rule-based frameworks where procedural justice is important; and decisions made under cognitive impairment or vulnerability, for example by elderly decision-makers.

This project is part of the Agentic AI Research Group within WASP-HS (see below for a description of WASP-HS). The group focuses specifically on agentic AI: systems capable of increasingly autonomous planning, coordination, and action. It studies such systems not only as a societal and scientific challenge, but also as potential collaborators in the research itself. The work therefore revolves around human-AI collaboration, alignment, and the development of new research practices where autonomous AI systems can contribute to the generation, testing, and evaluation of knowledge.

The postdoc will be based at Lund University Cognitive Science and work in close collaboration with members of the Agentic AI Research Group. Current core members of the group are Christian Balkenius (LUCS), Annika Wallin (LUCS), Petter Johansson (LUCS), Lars Hall (LUCS), Daniel Västfjäll (Linköping University), Philip Pärnamets (Karolinska Institutet), Anna Dreber (Stockholm School of Economics).

Wallenberg AI, Autonomous Systems and Software Program – Humanity and Society (WASP-HS) is a national research program in Sweden. The vision for WASP-HS is to promote new interdisciplinary knowledge within the humanities and social sciences regarding AI and autonomous systems and their impact on human and societal development. WASP-HS enables groundbreaking research, expertise, and competence building within the humanities and social sciences. https://wasp-hs.org/ (https://wasp-hs.org/)

Duties and Responsibilities

As a postdoc, you will play a central role in developing and conducting research on emotionally capable LLMs and human-AI collaboration. You will work with the entire research process, including theory development, study design, data collection, statistical analysis, interpretation, and publication. You will design and conduct experiments on human interaction with artificial agents, contribute to modeling affective processes in AI systems, and investigate how emotional signaling can improve trust, collaboration, and decision quality in teams consisting of humans and AI.

You will also contribute to linking the project to current developments in agentic AI. This means you are expected to stay updated on research regarding LLM-based agents, prompting methods, fine-tuning strategies, and multi-agent systems, as well as assess how such methods can be integrated into the project's empirical and theoretical work. The position is therefore particularly well-suited for a candidate who wishes to combine strong experimental research skills with active engagement in emerging AI methods.

Qualifications

Requirements for the position:

• A PhD or equivalent foreign degree in cognitive science, psychology, affective neuroscience, human-computer interaction, or a related field. The PhD must be completed no later than the time the employment decision is made, and the applicant must complete their application with documentation proving the awarded PhD before this time.
• Good ability to develop and conduct high-quality research.
• Pedagogical ability.
• You must not have previously been employed as a postdoc supported by the current collective agreement for more than one year within the same or a closely related subject area at Lund University.
• Very good knowledge of English.
• Good understanding of central theories in affective science and social cognition (e.g., appraisal theories, basic vs. complex emotions, cognitive vs. affective empathy, shared intentionality, emotion regulation).
• Documented interest in, or previous work on, emotional capabilities of LLMs or other AI systems.
• Programming skills in Python sufficient for experimental work and data analysis.

Merits for the position:

• PhD awarded no more than three years before the application deadline. If there are special reasons such as illness, parental leave, union trust positions, service in the total defense, or service/duties relevant to the subject area, this time may be deducted from the time elapsed since the PhD was awarded.
• Experience in building or fine-tuning LLMs, for example through prompt engineering, LoRA (Low-Rank Adaptation), RLHF (Reinforcement Learning from Human Feedback), or full fine-tuning.
• Documented knowledge of affective computing or biologically inspired emotion models.
• Previous work with multi-agent simulations using LLMs on social behavior or social dilemmas.
• Knowledge of reinforcement learning applied to language agents or emotional decision-making.
• Competence in advanced prompting techniques (Chain-of-Thought, Tree-of-Thought, self-consistency, emotion-conditioned generation, and related techniques).
• Documented experience of interdisciplinary collaboration between cognitive science/psychology and AI/ML teams.