Explanation and inference: mechanistic and functional explanations guide property generalization

By Tania Lombrozo, Nicholas Z. Gwynne. 2014

58 citations recorded by the source corpus.

The ability to generalize from the known to the unknown is central to learning and inference. Two experiments explore the relationship between how a property is explained and how that property is generalized to novel species and artifacts. The experiments contrast the consequences of explaining a property mechanistically, by appeal to parts and processes, with the consequences of explaining the property functionally, by appeal to functions and goals. The findings suggest that properties that are explained functionally are more likely to be generalized on the basis of shared functions, with a weaker relationship between mechanistic explanations and generalization on the basis of shared parts and processes. The influence of explanation type on generalization holds even though all participants are provided with the same mechanistic and functional information, and whether an explanation type is freely generated (Experiment 1), experimentally provided (Experiment 2), or experimentally induced (Experiment 2). The experiments also demonstrate that explanations and generalizations of a particular type (mechanistic or functional) can be experimentally induced by providing sample explanations of that type, with a comparable effect when the sample explanations come from the same domain or from a different domains. These results suggest that explanations serve as a guide to generalization, and contribute to a growing body of work supporting the value of distinguishing mechanistic and functional explanations.

Subfields: Artificial Intelligence, History and Philosophy of Science, Molecular Biology

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In History and Philosophy of Science

In Molecular Biology

Citations in the corpus

Explanation in Human-AI Systems: A Literature Meta-Review, Synopsis of Key Ideas and Publications, and Bibliography for Explainable AI (2019, background): “…Rozenblit, and Mills, 2004; Koehler, 1991; Kuhn, 2001; Leddo, Abelson, and Gross, 1984; Lombrozo, 2006, 2009a, 2011, 2016; Lombrozo and Carey, 2006; Lombrozo and Gwynne, 2014a; Lombrozo, Sloman, Strevens, Trout, and Skolnick Weisberg, 2008; Malle, 1999,2004, 2011; Manicas and Secord, 1983;…”
Explanation in Artificial Intelligence: Insights from the Social Sciences (2017, background): “Harman [62] is perhaps the first to acknowledge this link, and more recently, experimental evaluations have demonstrated it [108, 187, 109, 153].”
Explanations in the wild. (2021, unknown): “The relationship between these two content features is likely complex (Liquin & Lombrozo, 2018; Lombrozo & Gwynne, 2014; McCarthy & Keil, 2022) but an ‘explanations in the wild’ approach – as a complement to traditional lab studies – can further help identify the circumstances in which causation…”
The scope of formal explanation (2017, background): “In fact, our conceptual systems naturally make use of multiple modes of explanation (e.g., Keil, 1994; Kelemen, 1999; Kelemen & Rosset, 2009; Lombrozo, 2009; Lombrozo & Gwynne, 2014; Prasada & Dillingham, 2006).”