Mechanism of duration perception in artificial brains suggests new model of attentional entrainment

Abstract

While cognitive theory has advanced several candidate frameworks to explain attentional entrainment, the neural basis for the temporal allocation of attention is unknown. Here we present a new model of attentional entrainment that is guided by empirical evidence obtained using a cohort of 50 artificial brains. These brains were evolved in silico to perform a duration judgement task similar to one where human subjects perform duration judgements in auditory oddball paradigms. We found that the artificial brains display psychometric characteristics remarkably similar to those of human listeners, and also exhibit similar patterns of distortions of perception when presented with out-of-rhythm oddballs. A detailed analysis of mechanisms behind the duration distortion in the artificial brains suggests that their attention peaks at the end of the tone, which is inconsistent with previous attentional entrainment models. Instead, our extended model of entrainment emphasises increased attention to those aspects of the stimulus that the brain expects to be highly informative.

Psychophysical Tests Reveal that Evolved Artificial Brains Perceive Time like Humans

“…Just as observed with the human subjects (McAuley and Fromboluti, 2014), the digital brains misjudge the length of these tones (see Fig. 4), with the error more pronounced the more off-rhythm the oddball is (Tehrani-Saleh et al, 2020).…”

“…Because digital brains can be examined with perfect precision, the mechanisms behind the observed perceptual biases can be determined without relying on inference, to discover new theories of cognitive processing. For example, in (Tehrani-Saleh et al, 2020) we determined that the distortion in duration perception is due to brains only paying attention to the most informative parts of the signal (in this case, the endpoint), suggesting a new theory of attention.…”

“…For these experiments (Tehrani-Saleh et al, 2020) we are able to vary tone length and IOI to a much larger extent compared with human subjects. Using a unit tone length corresponding to 50msec per unit, we evolved Markov Brains with IOIs ranging from 10 to 25 (with a standard tone length about half the IOI) and all possible tone lengths in between the IOI for the oddball.…”

Using MAP-Elites to direct the evolution of desired neural characteristics

“…However, many of the researchers that were spearheading the "Deep Learning" revolution now concede that those networks are vulnerable to so-called "adversarial perturbations" (Jo and Bengio, 2018), and are unlikely to represent a step towards artificial general intelligence, because they do not develop robust mental models. gramming mental representations, it is possible to produce them by evolving artificial brains that control agents that act and behave in complex simulated environments (Edlund et al, 2011;Marstaller et al, 2013;Albantakis et al, 2014;Tehrani-Saleh et al, 2018;Olson et al, 2016;Tehrani-Saleh et al, 2019). Within these digital brains (Markov Brains, see Hintze et al (2017)) mental representations are used to make decisions in conjunction (and sometimes even without) reference to sensory information.…”