Arturo Vazquez
University of Southampton
The relationship between psychological language and artificial behaviour remains puzzling in current debates on the philosophy of mind. This paper aims to offer a plausible account of a specific use of psychological language in this context. When programming a machine learning algorithm to solve a particular task, we might employ empirical, truth-bearer propositions to report its behaviour, for example, ‘the machine did not recognise enough cat-images’, ‘… is currently calculating the most probable outcome’, and ‘… checkmated the chess engine’.
Given that (i) we sometimes describe the behavioural patterns of artificial systems in perceptual, intentional, normative, or dispositional terms, and (ii) the nature of these systems is substantially different from what ordinarily displays mental capacities, it makes sense to ask what the status of this employment of language is. In other words, what we mean when attributing psychological predicates to machines when describing their behaviour patterns is yet to be clarified — this is the ‘Problem of Artificial Behaviour’ (PAB).
Section 1 elaborates on PAB and its different variations (e.g. ‘What does it mean to say that a machine identifies a particular pattern or calculates the result of a mathematical operation?’). Section 2 shows that a Cartesian notion of behaviour does not provide an answer to PAB, as a ‘non-variational’ view of language makes the relation between psychological terms and artificial behaviour opaque. In section 3, we expand on a meaningful conception of behaviour and argue that such conception helps clarify PAB as this alternative notion displays two features — (i) a ‘variational’ view of language and (ii) a strong notion of ‘agreement’ — that account for a descriptive employment of language in the context of artificial behaviour.
Chair: Mathilde Cappelli
Time: September 12th, 10:40 – 11:10
Location: SR 1.006 (online)
SOPhiA 2026 