Affective Brain Group

Head: Tali Sharot

Information-seeking as a window to mental health. We have previously theorized (Sharot & Sunstein, Nature Human Behaviour, 2020) that when seeking information people estimate the expected impact of information on their actions (will the knowledge help or hinder my ability to make good decisions?), affect (will the knowledge induce positive or negative feelings?) and cognition (how will the knowledge alter my understanding of the world?). These estimates are integrated into a computation of the value of information, triggering either information-seeking or its avoidance. We have shown that our model is successful in explaining information-seeking behaviour, performing better than other models (Kelly & Sharot, Nature Communication, 2021; Cogliati Dezza, Maher, & Sharot, Cognition ,2022). Different individuals assign different weights to these three factors (e.g., some care a lot about how information will impact their emotions, others do not), and these weights are relatively stable over time and domain, providing clues about mental health (Kelly & Sharot, Nature Communication, 2021).

We are currently developing tools to extract markers from web browsing that provide indicators of mental health. As part of this endeavor, we are developing algorithms to quantify features of the text people consume online (e.g., affective sentiment, instrumental utility, semantic relatedness between searches). Our early results reveal a replicable association between mental health and affective properties of text people consume from self-guided searches (Kelly & Sharot, in prep).

The nature of intrinsic rewards. We have suggested that the rewarding nature of almost all intrinsic rewards are due to two features: increased sense of self efficiency and increased sense of agency (Blain & Sharot, Current Opinion in Behavioural Science, 2021). We have further shown that the sensitivity to intrinsic rewards is domain general and related to affective health (Blain, Pinhorn & Sharot, in prep). We now aim to elucidate the common neural mechanism of intrinsic rewards.

AI-Human interaction. In 2021 we launched a series of studies to test how AI-human interactions alter human cognition. We find that after collaborating with a biased AI system, participants become more biased themselves - an effect they are unaware of. Conversely, after interacting with accurate AI systems, participants became more accurate. We are elucidating the mechanisms by which AI systems influence human judgement to understand how they are different from, or similar to, human-human influence. This knowledge can be used to improve AI-human interaction and reduce the potential for discriminatory impact.

Key publications

  1. Kelly C.A. & Sharot, T. (2021) Individual differences in information-seeking. Nature Communication, 12, 7062.
  2. Vellani, V., de Vries, L.P., Gaule, A., Sharot, T. (2020). A selective effect of dopamine on information-seeking. eLife, 9, e59152.
  3. Kappes, A., Harvey, A., Lohrenz, T., Montague, R. Sharot, T. (2020) Confirmation Bias in the Utilization of Others’ Opinion Strength. Nature Neuroscience, 23(1), 130-137.
Go to Editor View