van Lier, M. (2023). Introducing a four-fold way to conceptualize artificial agency. Synthese, 201(85). doi:10.1007/s11229-023-04083-9
[Abstract]Recent developments in AI-research suggest that an AI-driven science might not be that far off. The research of [for] Melnikov et al. (2018) and that of Evans et al. (2018) show that automated systems can already have a distinctive role in the design of experiments and in directing future research. Common practice in many of the papers devoted to the automation of basic research is to refer to these automated systems as ‘agents’. What is this attribution of agency based on and to what extent is this an important notion in the broader context of an AI-driven science? In an attempt to answer these questions, this paper proposes a new methodological framework, introduced as the Four-Fold Framework, that can be used to conceptualize artificial agency in basic research. It consists of four modeling strategies, three of which were already identified and used by Sarkia (2021) to conceptualize ‘intentional agency’. The novelty of the framework is the inclusion of a fourth strategy, introduced as conceptual modeling, that adds a semantic dimension to the overall conceptualization. The strategy connects to the other strategies by modeling both the actual use of ‘artificial agency’ in basic research as well as what is meant by it in each of the other three strategies. This enables researchers to bridge the gap between theory and practice by comparing the meaning of artificial agency in both an academic as well as in a practical context.
[Citing Place (1956)]  
Citing Place (1956) in context (citations start with an asterisk *):
Section 3 The neo-Gricean framework
Subsection 3.3 Theoretical modeling and its implementation
* According to Sarkia, one example of theoretical modeling in the philosophy of mind and action is common sense or ‘analytical’ functionalism. This approach can be described as theoretical modeling, because it “involves indirect. representation of agents through general law-like regularities connecting mental states to behavior” (§4). This kind of functionalism was initially meant as a response to the Psycho-Physical Identity Theory (see e.g. Place 1956; Feigl 1958; Smart 1959; Hill 1991; Polger 2011), which holds that mental states correspond to particular physical brain states. In functionalism, these mental states instead correspond to the causal role that they play in the cognitive system that they are part of, which enables researchers to abstract away from particular brain states. By focusing on this causal role that they play in a cognitive system, mental states become multiple realizable, meaning that one particular mental state can be realized by different (brain) states as long as the mental state still plays the same causal role in the cognitive system. This multiple realizability is what makes functionalism in the philosophy of mind and action interesting for accounts of artificial agency as well.