Yamaguchi, Mako and Sugio, Hajime (2025) Computationally Reframing the Theory-Ladenness of Observation: Hanson’s Theory-Ladenness, Predictive Processing, and the Bayesian Structure of Scientific Discovery. [Preprint]
![[img]](https://philsci-archive.pitt.edu/style/images/fileicons/text.png) |
Text
Hanson_MakoHajime.pdf Download (640kB) |
Abstract
This paper offers a computationally explicit reformulation of N. R. Hanson’s thesis that observation is theory-laden seeing-as. Drawing on predictive processing and Bayesian modeling, it advances a failure-sensitive bridge hypothesis: a role- and constraint-preserving correspondence—not a reductive identification—between personal-level philosophical roles and sub-personal computational roles.
Crucially, the paper rejects the slogan “theory = prior.” Instead, it models theories as meta-level admissibility constraints on model space. These constraints determine which variables, structures, measurement assumptions, and relatively durable precision expectations are admissible. On this view, what can count as an observation report or a scientific “fact” is shaped by inference and by policy-governed reporting under theory-conditioned constraints, together with stabilized measurement and reporting regimes.
The first part reconstructs Hanson’s analysis in computational terms. Seeing-as is modeled as posterior-guided inference under a theory-conditioned generative model, while observation statements are treated as goal-, context-, and norm-sensitive report policies over posterior states rather than as direct readouts of raw data. The second part recasts discovery dynamics as multi-tier revision. Apparent mismatches may be resolved by within-model routes such as parameter retuning, measurement or noise-model repair, and precision reallocation. By contrast, persistent structured residuals—stable across plausible measurement and noise variants—can rationally motivate between-model selection and, when required, revision of the admissible model space itself. A reanalysis of Kepler’s adoption of elliptical orbits illustrates how such constraint changes reorganize the space of admissible observations, facts, and explanations.
Overall, the paper proposes a computationally explicit yet non-reductive framework that clarifies the structural relations among theory, perception, and scientific discovery.
| Export/Citation: | EndNote | BibTeX | Dublin Core | ASCII/Text Citation (Chicago) | HTML Citation | OpenURL |
| Social Networking: |
| Item Type: | Preprint | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Creators: |
|
|||||||||
| Keywords: | Theory-ladenness of observation; Seeing-as; Predictive processing; Bayesian modeling; Hierarchical generative models; Precision weighting; Admissibility constraints on model space; Report policies; Model evidence and model comparison; Abduction (inference to the best explanation); Scientific discovery and theory change | |||||||||
| Subjects: | Specific Sciences > Cognitive Science > Computation Specific Sciences > Cognitive Science > Concepts and Representations Specific Sciences > Cognitive Science > Perception General Issues > Philosophers of Science |
|||||||||
| Depositing User: | Mako Yamaguchi | |||||||||
| Date Deposited: | 22 Dec 2025 13:39 | |||||||||
| Last Modified: | 22 Dec 2025 13:39 | |||||||||
| Item ID: | 27598 | |||||||||
| Subjects: | Specific Sciences > Cognitive Science > Computation Specific Sciences > Cognitive Science > Concepts and Representations Specific Sciences > Cognitive Science > Perception General Issues > Philosophers of Science |
|||||||||
| Date: | 21 December 2025 | |||||||||
| URI: | https://philsci-archive.pitt.edu/id/eprint/27598 |
Monthly Views for the past 3 years
Monthly Downloads for the past 3 years
Plum Analytics
Actions (login required)
 |
View Item |
![[feed]](http://philsci-archive.pitt.edu/style/images/feed-icon-32x32.png){kind=icon}