Closing the gap between ideal and real behavior: Scientific vs. engineering approaches to normativity

Abstract

Early normative studies of human behavior revealed a gap between the norms of practical rationality (what humans ought to do) and the actual human behavior (what they do). It has been suggested that, to close the gap between the descriptive and the normative, one has to revise norms of practical rationality according to the Quinean, engineering view of normativity. On this view, the norms must be designed such that they effectively account for behavior. I review recent studies of human perception which pursued normative modeling and which found good agreement between the normative prescriptions and the actual behavior. I make the case that the goals and methods of this work have been incompatible with those of the engineering approach. I argue that norms of perception and action are observer-independent properties of biological agents; the norms are discovered using methods of natural sciences rather than the norms are designed to fit the observed behavior.

Keywords:

• Bayesian
• Decision
• Economics
• Engineering
• Haptics
• Illusion
• Normativity
• Psychophysics
• Rationality
• Vision

Acknowledgements

I wish to thank John Jacobson, Michael Kubovy, and Cees von Leeuwen for valuable comments, and the National Institute of Natural Sciences of Japan and the Swartz Foundation for generous financial support.

Notes

Notes

[1] One reason visual measurement are inherently ambiguous is the optical projection from the three-dimensional scenes to the effectively two-dimensional retinal surface, such that the same retinal image may correspond to different stimuli. An example of noisy biological computation is the absorption of light by retinal receptors, which is a stochastic process.

[2] This knowledge can be implicit, e.g., implemented in automatic neural computations, possibly inaccessible to awareness. This knowledge is usually described by saying that nervous systems “take into account” or “represent” regularities of the environment.

[3] The notion of reliability has a technical meaning; it is defined as the inverse of the variance of the distribution of estimates. Precision of estimation is defined as the standard deviation of this distribution.

[4] The notion of “ideal observer” can be used narrowly or broadly. In the narrow sense, ideal observer models disregard the constraints of biological computation, so the researcher may compare performance of a biological system with a mathematical ideal. In the broad sense, ideal observer models incorporate some biological constraints, such as the decision noise in the Statistical Decision Theory (Geisler, 1989; Green & Swets, 1966). Here I use the latter approach.

[5] “Haptic” information is sensory information gathered through active touch, using tactile and proprioceptive signals.

[6] For example, touch is more reliable than vision for estimating roughness of surfaces (Lederman & Abbott, 1981).

[7] This is not to say that the cue-integration models and the Bayesian inference models are incompatible. On the contrary, the two kinds of models belong to the same decision-theoretic framework and are readily combined (e.g., Hillis, Watt, Landy, & Banks, 2004).