The minority report: some common assumptions to reconsider in the modelling of the brain and behaviour

Abstract

Reverse-engineering the brain involves adopting and testing a hierarchy of working hypotheses regarding the computational problems that it solves, the representations and algorithms that it employs and the manner in which these are implemented. Because problem-level assumptions set the course for the entire research programme, it is particularly important to be open to the possibility that we have them wrong, but tacit algorithm-and implementation-level hypotheses can also benefit from occasional scrutiny. This paper focuses on the extent to which our computational understanding of how the brain works is shaped by three such rarely discussed assumptions, which span the levels of Marr's hierarchy: (i) that animal behaviour amounts to a series of stimulus/response bouts, (ii) that learning can be adequately modelled as being driven by the optimisation of a fixed objective function and (iii) that massively parallel, uniformly connected layered or recurrent network architectures suffice to support learning and behaviour. In comparison, a more realistic approach acknowledges that animal behaviour in the wild is characterised by dynamically branching serial order and is often agentic rather than reactive. Arguably, such behaviour calls for open-ended learning of world structure and may require a neural architecture that includes precisely wired circuits reflecting the serial and branching structure of behavioural tasks.

The problem of multimodal concurrent serial order in behavior

“…The LSC framework includes both formal semantics and software development tools and is based on "multi-modal scenarios, each corresponding to an individual requirement, specifying what can, must, or may not happen following certain sequences of events" A c c e p t e d M a n u s c r i p t (Harel, Marron, and Weiss, 2012). LSC is a dual formalism to that of statecharts (Harel, 1988(Harel, , 2007, which came up as a candidate formalism in our earlier work (Edelman, 2011;Goldstein et al, 2010;Kolodny et al, 2015;Edelman, 2015b). Whereas statecharts focus on states that a system can occupy (including a superposition or Cartesian product of several states at once), Live Sequence Charts focus on processes and events.…”

“…In a recent review, Edelman (2015b) documented the pervasiveness of the stimulus/response doctrine, noting that its resilience is particularly surprising, given that it had been considered problematic already over a century ago, when John Dewey first offered a critique of "the reflex arc concept in psychology": "What we have is a circuit, not an arc or broken segment of a circle. [.…”

“…While they do serve as an essential basis for behavior in all vertebrates, in mammals and birds these representations are augmented with information mediated, respectively, by the isocortex and by the corresponding parts of the pallium. This information includes hierarchically structured sensorimotor representations, carried by occipito-temporal and parietal pathways, and representations supporting planning and problem solving, likewise hierarchical, in the frontal lobe (see Edelman, 2015b and the many references therein for a review). The allocortical structure where these pathways converge is the hippocampus (Buzsáki, 1996;Merker, 2004).…”

The evolution of cognitive mechanisms in response to cultural innovations

“…Instead, we want a minimal set of principles that suffice to explain how simple operational units, capable of only the most basic forms of learning, can jointly and gradually create the much more sophisticated mechanisms of advanced cognition. [Powerful algorithms using deep learning (76,77), with less emphasis on biological and behavioral realism (78), are being developed and applied to challenging tasks, including: perceptual parsing, associative learning, and the learning of conceptual contingencies (79,80). ]…”

Constructive anthropomorphism: a functional evolutionary approach to the study of human-like cognitive mechanisms in animals

“…Finally, relying on similar theoretical approaches, we assume that cognitive mechanisms (including highly advanced ones in humans), can be broken down into associative learning principles that can construct complex representations of past experiences in the brain [17,[27][28][29][30], most probably in the form of a network [31][32][33][34]. Accordingly, advanced cognitive mechanisms are not viewed as alternatives to associative learning but rather as mechanisms that evolved from, and are based on, associative principles.…”