Fabian Grabenhorst’s research while affiliated with University of Oxford and other places

Saving behavior usually require individuals to perform several consecutive choices before collecting the final reward. The overt behavior is preceded by an intention to perform an appropriate choice sequence. We studied saving sequences for which each participant rated the intention numerically as willingness-to-save. Each sequence resulted in a specific reward amount and thus had a particular value for the participant, which we assessed with a Becker-DeGroot-Marschak (BDM) auction-like mechanism. Using functional MRI, we found that blood oxygen level dependent (BOLD) signals in human ventromedial prefrontal cortex (vmPFC) correlated with the participant's stated intention before each choice sequence. An adjacent vmPFC region showed graded activation that reflected the value of the sequence. These results demonstrate an involvement of vmPFC in intentional processes preceding sequential economic choices.

Realistic, everyday rewards contain multiple components. An apple has taste and size. However, we choose in single dimensions, simply preferring some apples to others. How can such single-dimensional relationships refer to multi-dimensional choice options? Here, we investigated stochastic choices of two-component milkshakes. The revealed preferences were intuitively graphed as indifference curves that indicated how much of one component was traded-in for obtaining one unit of the other component without a change in preference, thus defining the orderly integration of multiple components into single-dimensional estimates. Options on higher indifference curves were preferred to those on lower curves. The systematic, non-overlapping curves satisfied leave-one-out tests, followed decoder predictions and correlated with Becker-DeGroot-Marschak auction-like bids. These single-dimensional estimates of multi-component options complied with rigorous concepts of Revealed Preference Theory and encourage formal investigations of normal, irrational and pathological decisions and their neural signals.

Risk derives from the variation of rewards and governs economic decisions, yet how the brain calculates risk from the frequency of experienced events, rather than from explicit risk-descriptive cues, remains unclear. Here, we investigated whether neurons in dorsolateral prefrontal cortex process risk derived from reward experience. Monkeys performed in a probabilistic choice task in which the statistical variance of experienced rewards evolved continually. During these choices, prefrontal neurons signaled the reward-variance associated with specific objects ('object risk') or actions ('action risk'). Crucially, risk was not derived from explicit, risk-descriptive cues but calculated internally from the variance of recently experienced rewards. Support-vector-machine decoding demonstrated accurate neuronal risk discrimination. Within trials, neuronal signals transitioned from experienced reward to risk (risk updating) and from risk to upcoming choice (choice computation). Thus, prefrontal neurons encode the statistical variance of recently experienced rewards, complying with formal decision variables of object risk and action risk.

Artificial agents are becoming prevalent across human life domains. However, the neural mechanisms underlying human responses to these new, artificial social partners remain unclear. The uncanny valley (UV) hypothesis predicts that humans prefer anthropomorphic agents but reject them if they become too humanlike-the so-called UV reaction. Using fMRI, we investigated neural activity when subjects evaluated artificial agents and made decisions about them. Across two experimental tasks, the ventromedial prefrontal cortex (VMPFC) encoded an explicit representation of subjects' UV reactions. Specifically, VMPFC signaled the subjective likability of artificial agents as a nonlinear function of humanlikeness, with selective low likability for highly humanlike agents. In exploratory across-subject analyses, these effects explained individual differences in psychophysical evaluations and preference choices. Functionally connected areas encoded critical inputs for these signals: the temporoparietal junction encoded a linear humanlikeness continuum, whereas nonlinear representations of humanlikeness in dorsomedial prefrontal cortex (DMPFC) and fusiform gyrus emphasized a human-nonhuman distinction. Following principles of multisensory integration, multiplicative combination of these signals reconstructed VMPFC's valuation function. During decision making, separate signals in VMPFC and DMPFC encoded subjects' decision variable for choices involving humans or artificial agents, respectively. A distinct amygdala signal predicted rejection of artificial agents. Our data suggest that human reactions toward artificial agents are governed by a neural mechanism that generates a selective, nonlinear valuation in response to a specific feature combination (humanlikeness in nonhuman agents). Thus, a basic principle known from sensory coding-neural feature selectivity from linear-nonlinear transformation-may also underlie human responses to artificial social partners.SIGNIFICANCE STATEMENT Would you trust a robot to make decisions for you? Autonomous artificial agents are increasingly entering our lives, but how the human brain responds to these new artificial social partners remains unclear. The uncanny valley (UV) hypothesis-an influential psychological framework-captures the observation that human responses to artificial agents are nonlinear: we like increasingly anthropomorphic artificial agents, but feel uncomfortable if they become too humanlike. Here we investigated neural activity when humans evaluated artificial agents and made personal decisions about them. Our findings suggest a novel neurobiological conceptualization of human responses toward artificial agents: the UV reaction-a selective dislike of highly humanlike agents-is based on nonlinear value-coding in ventromedial prefrontal cortex, a key component of the brain's reward system.

By observing their social partners, primates learn about reward values of objects. Here, we show that monkeys’ amygdala neurons derive object values from observation and use these values to simulate a partner monkey’s decision process. While monkeys alternated making reward-based choices, amygdala neurons encoded object-specific values learned from observation. Dynamic activities converted these values to representations of the recorded monkey’s own choices. Surprisingly, the same activity patterns unfolded spontaneously before partner’s choices in separate neurons, as if these neurons simulated the partner’s decision-making. These ‘‘simulation neurons’’ encoded signatures of mutual-inhibitory decision computation, including value comparisons and value-to-choice conversions, resulting in accurate predictions of partner’s choices. Population decoding identified differential contributions of amygdala subnuclei. Biophysical modeling of amygdala circuits showed that simulation neurons emerge naturally from convergence between object-value neurons and self-other neurons. By simulating decision computations during observation, these neurons could allow primates to reconstruct their social partners’ mental states.

This corrects the article DOI: 10.1038/ncomms12554.