Schrager, J. (1990). Commonsense perception and the psychology of theory formation. In Shrager, J. & Langley, P. (Eds.) Computational Models of Scientific Discovery and Theory Formation. Morgan Kaufman, San Mateo, CA. 437--470.

@InBook{,
  ALTauthor = 	 {Schrager, J.},
  ALTeditor = 	 {Schrager, J. and Langley, P.},
  title = 	 {Computational
Models of Scientific Discovery and Theory Formation},
  chapter = 	 {Commonsense perception and the psychology of
theory formation},
  publisher = 	 {Morgan Kaufman},
  year = 	 {1990},
  OPTpages = 	 {437--470},
}

Author of the summary: Jim Davies, 2002, jim@jimdavies.org

Cite this paper for:

• "Reinterpretation plays a central role in theory formation."
• "Sensory context is a fundamental and inextricable aspect of the theories formed and their formation by reinterpretation."
• Commonsense perception is a central psychological mechanism that plays a key role in both theory formation and complex reasoning.
• Quasi-linguistic: representations that approximate english words of phrases [348]
• view application
• Grounded representation: Sensation and sensory experience form the basis of conceptual knowledge [450]
• paradox of recognition [448]
• bitmap representations

Quasi-linguistic: representations that approximate english words of phrases [348]

Schrager and Klahr (1983, 1986): Gave Ss BigTrak toys and asked them to figure it out. They would try something, and when the observed behavior did not jibe with expectations, they recognized what was actually happening (vector addition, e.g.) and put that concept into their theories. Steps are 1. recognition, 2. Introduction of a concept into a theory, and 3. Reformulation of the theory. [440]

Schrager and Klahr (1983, 1986): In the Digicomp example (a flip flop and straw mechanical computer) Ss created a new vocabulary (e.g. blocking, setting) for themselves as they learned about the device.

These processes are called "view application." [447] viewing the device under study as an instance of a commonsense abstraction, using incremental conceptual combination of "views" (abstract schemas) and their present theory. There is view selection and combination. Selection is a recognition.

Grounded representation: Sensation and sensory experience form the basis of conceptual knowledge [450].

The paradox of recognition is this: how can you recognize something as an example of something else if the two vocabularies don't match? The solution is that representations of both have a more primitive level found in experience traces. [451]

The computational agent system has list representations and a 100x100 pixel iconic store. It contains tokens representing the objects understood to be in that location. [457]

Visual primitives: Circle [458], left-of, radius

Radius is an abstraction for excitation of the molecule [459].

The agent uses a series of bitmapped snapshots. Processes are recognized by image matching. [461]

Summary author's notes:

• They use the word "modality" to mean the format part of representation. So ACT-R and SME would be theories at the modality level.
• The recognition paradox is the symbolic mismatch problem.