Online Lecture by Anne Urai
Shared decision computations underlie choice behavior in mice and humans
Recent advances in training mice to perform complex tasks, combined with powerful optogenetic and neural measurement tools, have positioned mice as an ideal model species for probing the neural circuit mechanisms of cognition. An important assumption underlying this line of work is that these mechanisms are preserved across mammalian species, and provide insight into the same cognitive processes that play out in the human brain. However, this assumption is rarely explicitly tested, creating challenges in the translatability of neuroscience findings from mice to humans. We here test directly if humans and mice use the same computations to perform sensory-guided decision-making. We focus on a signature of evidence accumulation over multiple temporal scales, that robustly captures individual behavioral differences in humans across a range of tasks: a change in the dynamics of evidence integration that depends on the agent’s previous choices (Urai et al. 2019). Using data from a large-scale neuroscience collaboration (The International Brain Laboratory et al., 2020), we analyzed data from 100 mice that learned to perform a visual decision-making task. Applying the same drift diffusion model (Wiecki et al. 2013) as in humans, we found that choice history biases were explained by the same computational principle: a history-dependent change in the rate of evidence accumulation. These findings suggest that evidence accumulation over multiple temporal scales reflect a fundamental aspect of decision-making, that may be conserved across mammalian species. More importantly, this allows us to hypothesize that the neural circuit mechanisms that give rise to choice history behavior in mice are relevant for understanding the human brain.