The putative emergence of emotions from sensory systems: from William James to artificial moral agents (AMAs)

After acknowledging that artificial intelligence engineers “are a long way off from knowing how to develop systems that can feel pleasure or pain, or have human-like emotions,”* Wendell Wallach and Colin Allen are, nonetheless, no less hopeful for the prospects of progress on this front: “[S]ensory technology is an active area of research, and it is here that one might look for the foundations of feelings and emotions” in AI and robotics.

So, the pivotal assumption is that sensory modalities provide the (causal) foundations of feelings and emotions and, in principle, we can construct technologies somehow capable of replicating these animal and human modalities which will thus get us that much closer to developing the requisite AI technologies in possession of the ability to “feel pleasure or pain, or have human-like emotions.” This is, in effect, at once both a reductionist and emergentist model (‘reductionist’ insofar as it traces feelings and emotions back to the senses, and ‘emergentist’ insofar as the sensory modalities provide us with the ‘data’ which will make possible the construction of AI/robotic systems that feel pleasure and pain as well as have ‘human-like emotions’) of the emotions that is no less troubling than (related if not similar) reductionist accounts of the mind. And it has scientific pedigree back to the nineteenth century in the “William James-C. Lange psychological theory of the emotions” which, in turn, has of late been resurrected in the “somatic marker theory” provided by the cognitive neuroscientist Antonio Damasio. Damasio’s theory has been accorded pride of place among most twenty-first century neuroscientists (and not a few scientistically-inclined philosophers). As P.M.S. Hacker argues, ‘[b]oth theories have striking similarities to Descartes’s” and all three “vividly demonstrate how failure to attend adequately to conceptual questions concerning the nature of the emotions leads to conceptual confusions in the construction of what purport to be empirical theories of the emotions.”

Not surprisingly, then, our co-authors take on board the philosophical and psychological confusions that plague James’s theory and re-appear in the main with Damasio. Hacker attends to the fundamental conceptual questions avoided by these empirical theories, which allows him to proffer a powerful and succinct critique of these accounts in his latest book, The Passions: A Study in Human Nature (John Wiley & Sons, 2018).

Wallach and Allen subscribe to a bodily-based “building-blocks”/emergentist model of the emotions that Hacker sufficiently—albeit indirectly—dismantles with his critique (ad hoc additions from evolutionary psychology in no way improve this model’s plausibility) of James and Damasio. They concede that “sensory technology” has not, as yet, provided the aforementioned “foundations” (i.e. ‘building-blocks’) for the emotions, but their subscription to this thoroughly discredited empirical theory has not expired:

“These newer sensory technologies, combined with much older technologies of cameras and microphones, allow considerable amounts of sensory data to be accumulated. However, the next step—mapping these data onto the feelings and emotions that motivate actions [of course not all feelings and emotions are motivational in this way]—is more difficult. Emotions and other mental states [magically!] emerge from a web of inputs from different senses. [….] [C]omplex integrative somatic architectures [or, more plainly, bodily building-blocks] are clearly within the scope of artificial systems. It remains to be seen whether it is necessary to emulate in AMAs the full range of subtle emotional states evident in humans. The only way to find out is to build progressively more sophisticated system and test them in realistic situations.”

* I would argue that no such knowledge is forthcoming. For but a taste of the reasons why, please see this post: “A brief broadside on why AI systems or robots cannot—now or in the future—‘read’ our emotions.”