Responding on cue
Humans and animals alike are constantly engaged in making decisions based on visual or auditory cues from the environment. We have an extraordinary facility to make decisions based on our perceptions but are unable to make these decisions based on our cognitive abilities. The EU-funded project 'How well can humans perform: testing human cue integration across multiple systems' (Cue integration) is investigating the perceptive vs cognitive paradox, along with other critical characteristics for decision-making. This requires testing and modelling based on thoroughly researched data and computations. It involves a range of ideas touching on economics, psychology and related fields that have converged in a concept known as Bayesian optimal decision-making. To a large extent, when confronted with simple perceptual stimuli, humans and animals are frequently characterised as Bayes optimal, meaning they utilise and manipulate sensory information in a statistically optimal way. However, the non-Bayes optimal nature of cognitive decision-making has not been seriously studied in cognitive psychology and economics: the 1970s saw some testing on how humans simplify methods for problem solving to an extent that produces severely suboptimal behaviour. The Cue integration project is conducting a series of new model-based experiments to probe the continuum between perceptual and cognitive decision-making. It is identifying how subjects learn about cues and whether sub-optimal performance is based on inadequate learning, lack of evidence and/or other failures. The current study is examining the constraints on human cognitive decision making. In what way is this constrained, and is it possible with the right training, presentation and incentives for subjects to break these restrictions? This is a highly multidisciplinary study involving statistics, economics, and control theory to determine optimal decision-making and reinforcement learning to yield a more accurate model of learning. The Cue integration project team is using two different strategies to design the experiments. In one it is taking cognitive-prone tasks and adorning them with perceptual features; in the other it is taking perceptual-prone tasks and adding cognitive components. The team has conducted extensive experimental studies in both directions to see if performance could be improved. While the results were not as clear-cut as expected, the emerging data did provide interesting information as to how subjects solve such probabilistic tasks. More tests are ongoing to uncover the paradox of cue integration. These results may have the potential to change our understanding of human decision-making and unite research fields in novel ways.
