Habit-inspired methods for embodied intelligence

Abstract Algorithms that employ habitual processing may be characterized as context-contingent input-output pairings. They learn to relate recurring patterns with swift and effective responses, relying on minimal computational resources. While habits have been extensively researched in the fields of psychology, cognition, and neuroscience, they have yet to be systematically investigated in computer science. Data volumes are expanding at an exponential rate, and there is often a requirement for this data to be processed in real-time. Against this backdrop, principles of habitual cognition present a promising alternative to resource-intensive planning algorithms. We propose that future research focus on algorithmic and neuronal models of the neurocognitive foundations of habits to develop new perspectives on artificial intelligence and robotics.