Weinberger, Naftali (2023) Homeostasis and the Faithless Foundations of Causal Inference. [Preprint]

Preview

Text Homeostatic_Faithless_Foundations_ (9).pdf Download (2MB) | Preview

Abstract

The causal faithfulness condition, which licenses inferences from probabilistic to causal independence, is known to be violated in dynamical systems exhibiting homeostasis. Using the example of the Watt governor, I here present a precise causal characterization of such violations, which differ from cases involving cancelling paths. Traditional defenses of faithfulness do not carry over to cases of homeostasis and one should expect such cases to be widespread. Epistemically, this does not pose a problem for causal inference, since such failures of faithfulness do not undermine the ability of models at different time scales to properly represent the causal and probabilistic independence
relations effectively obtaining at the relevant time scale. These failures are nevertheless significant, since they provide the key to understanding how causal attributions can apply within complex systems exhibiting high degrees of mutual dependence among their parts.

---

Export/Citation:

EndNote | BibTeX | Dublin Core | ASCII/Text Citation (Chicago) | HTML Citation | OpenURL

Social Networking:

Share |

---

Item Type:	Preprint
Creators:	Creators Email ORCID Weinberger, Naftali naftali.weinberger@gmail.com
Keywords:	Causation; Dynamical Systems; Timescale; Complexity; Probability Theory;
Subjects:	General Issues > Causation Specific Sciences > Complex Systems Specific Sciences > Artificial Intelligence > Classical AI General Issues > Models and Idealization Specific Sciences > Probability/Statistics
Depositing User:	Mr. Naftali Weinberger
Date Deposited:	29 Jun 2023 13:24
Last Modified:	29 Jun 2023 13:24
Item ID:	22261
Subjects:	General Issues > Causation Specific Sciences > Complex Systems Specific Sciences > Artificial Intelligence > Classical AI General Issues > Models and Idealization Specific Sciences > Probability/Statistics
Date:	2023
URI:	https://philsci-archive.pitt.edu/id/eprint/22261

Monthly Views for the past 3 years

Monthly Downloads for the past 3 years

Plum Analytics

Actions (login required)

View Item