Supporting Meta-Cognition in LLM-based Exploratory Search
LLMs have so far demonstrated great potential in the enhancement of a vast variety of applications thanks to their adaptability, which allows them to integrate seamlessly into diverse scenarios. While such systems have impactfully augmented humans, they have also altered existing workflows in a way that can potentially disrupt human cognitive processes and lead to their progressive erosion. ; ; Exploratory Search activities are often characterized by vaguely defined and evolving information needs. The introduction of LLMs into such activities holds great potential for augmenting humans in performing information retrieval and summarization tasks. While in traditional search strategies users are encouraged to iteratively think through the refinement of their query, in LLM-based search, users receive a direct answer based on an initial, often underdeveloped, query, which may encourage them to bypass the critical process of iterative sensemaking and increase the risk of users accepting inaccurate or biased content. ; ; In this light, metacognition, defined as the ability of reflecting on one’s own thinking process and activating and utilizing certain cognitive abilities, is a key element in exploratory search activities as it helps users reflect on their information search strategies, recognize gaps in their understanding, and engage in a more intentional process of exploration. ; ; In this light, my doctoral research will focus on the development of a human-centered framework grounded in foundational theories of human cognition and information retrieval, such as Information Foraging Theory, designed for creating AI systems able to empower, rather than erode, human critical thinking.
Back to Current Students