THE FEEDBACK

Epistemic reduction of the concept of ‘decision’
By: Daniel Burnston (Tulane University)

Abstract: “Reduction” is a widely rejected view of how commonsense psychological notions relate to neuroscience. I argue that there is a particular view of reduction on which reduction of the key commonsense concept of ‘decision’ is a live option. In particular, I advance a version of epistemic reduction based on “connectability.” On this view, reduction relations are not logical relationships, but instead posited identifications that guide subsequent modeling efforts, and are substantiated if one can explain higher-level generalizations with lower-level models. I apply this view to the field of decision neuroscience, and in particular accumulation to bound models of decision-making. I show how this field is achieving the advantages of reductive explanation, including providing increasingly detailed and novel predictions, unifying distinct types of decisions under a single account, and explaining anomalies of rational decision-making. If I am right, the widespread anti-reductive attitude in philosophy of mind needs to be reconsidered.

Now published in Synthese.

Commentary from Daniel:
Most philosophy in recent decades has been guided, even if implicitly, by an anti-reductionist consensus.  On this view, psychological posits, especially those derived from our commonsense psychology, are autonomous from the neural sciences.  While there may be productive interplay between them, the nature of psychological states and processes is explained by psychological and philosophical theory, not neuroscientific results. 

I argue against this with two key moves.  The first is to adopt what I call “connectability-based reduction,” which stems from the work of Ernest Nagel.  On this view, rather than reduction requiring derivation of higher-level concepts from lower-level theories, reduction is a posited relation between concepts, undertaken in the course of explanation.  In essence, it is a commitment to model all of the generalizations surrounding the reduced concept with modeling resources derived from the reducing ones.  The second move is to endorse a model-based view of folk psychology, which argues that our commonsense psychological concepts are models (akin to folk physics) that we use to predict behavior.  As such, they can take part in scientific endeavor by structuring investigation, being reduced to, or being superseded by developed theoretical concepts and models. 

I argue that scientists use an intuitive sense of decision in approaching their investigations, and further that progress in neuroscience has involved the construction of a connectability relation between neuroscientific models and that commonsense notion.  The class of models I refer to as accumulation to bound models, which posit that ‘decision’ consists of accumulation of choice-relevant information towards a threshold, and the decision is implemented when the threshold is reached.  These kinds of processes are easily implemented by the brain through competitive dynamics and information integration, which are widespread brain functions.

In connectability-based reduction, the standards for success are explanatory rather than logical.  A posited connectability relation is supported when embracing it allows explanatory gains, by unifying instances of the reduced phenomena, explaining higher-level generalizations while also giving more detailed predictions, and by explaining anomalies.  I show how incredible progress has been made in all of these respects within the modeling field.  Accumulator models provide detailed predictions for such things as decision times, choice probabilities, and speed-accuracy tradeoffs, while also accounting for generalizations such as Hick’s law and frame effects.  They unify a range of decision phenomena across organisms, as well as across the many different kinds of decision that humans make.  And they given detailed explanations of “anomalies” of rational choice theories, such as failures of transitivity and preference reversals.

If I am right, then neuroscience is the field that explains the nature of decision.  And if that is right, then autonomy theses can no longer be maintained.  This result, especially given the centrality of decision to cognitive behavior, suggests that results from neuroscience can and should be integrated into discussions of agency, intention, control, and even moral responsibility.