Beyond Desire? Agency, Choice, and the Predictive Mind

ABSTRACT

‘Predictive Processing’ (PP) is an emerging paradigm in cognitive neuroscience that depicts the human mind as an uncertainty management system that constructs probabilistic predictions of sensory signals. Such accounts apply very naturally to perception and have plausible extensions to motor control. But desires and motivations can seem to pose a much greater puzzle, appearing especially resistant to reconstruction by a processing story that appeals to predictions alone. I examine several versions of this worry and show that it is fundamentally misplaced. Desires and motivations are fluently accommodated within the unifying PP schema, where they emerge as webs of prior ‘beliefs’ that sculpt probabilistic predictions, some of which become positioned (as we shall see) so as to bring about actions. Importantly, a single construct here plays the role of belief and desire. But what results is, perhaps surprisingly, a potentially richer landscape within which to think about agency, control, and choice.

KEYWORDS:

• desire
• belief
• prediction
• predictive processing
• agency
• control
• motivation

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Notes

1 This is by no means an ad hoc addition to the story. Recent PP-treatments of disorders of movement, such as of Parkinson’s disease, all point to empirically confirmed disruptions of sensory attenuation (see Brown, Adams et al. [2013] and Pareés et al [2014]; see also Palmer et al. [2016]).

2 More accurately, these comprise both probabilistic priors (when they are long-term or standing states) and predictions (when they are active states).

3 Here, and in the rest of this paragraph, I briefly rehearse the much richer and more detailed picture presented by Pezzulo et al. [2015, 2018].

4 I have not attempted to outline the PP account of other human attitudes here, but for some hints, see Joffily and Coricelli [2013].

5 Heartfelt thanks to the three anonymous referees whose careful and constructive comments have hugely improved this treatment. Thanks also to Anil Seth, Jakob Hohwy, Colin Klein, and Karl Friston for useful discussion of many of these ideas.

Additional information

Funding

The paper was written thanks to support from ERC Advanced Grant XSPECT—DLV—692739.