Friday, January 09, 2004

Perception as Inference

I just received an announcement of a conference about "visual thought."

Lovely expression!

Here is the announcement:

The depictive space of perception
A conference on visual thought

June, 7-9 2004
Mitteleuropa Foundation, Bolzano, Italy

Perceptual space and depictive space show strong similarities. Both are characterized by a sort of extendedness which unfolds dynamically, and which shows the close analogy between the performance of an act of perception and an act of design. Neither art nor vision are, in fact, veridical copies of the world, rather both seem to be operating on the representational structures of vision. On these premises, a scientific phenomenology, experimentally oriented, seems to be a more appropriate paradigm in vision science, especially in order to understand the dynamics of the ongoing perceiving. The conference has a starting point draws on the results of the artistic and cognitive theories of Klee and Arnheim and Gestalt theory, and explores their application to contemporary research in vision science.

Ample time will be allocated to discussion. If you are interested in attending the conference and/or contributing your own ideas, please send a mail (with a two-page abstract if you intend to give a paper) to Liliana Albertazzi ( before April, 15.

Invited Speakers
1. L. Albertazzi (Trento University), The Depictive Space of the Mind
2. C. E. Connor (John Hopkins University), Shape Representation in Neural Populations
3. T. Economou (Georgia Tech), Studies in Complexity, Ambiguity and Emergence in Design
4. F. Fol Leymarie (Brown University), The Computation of Visual Fields in Arts
5. J. Koenderink (Utrecht University), The Geometry of Pictorial Space
6. M. Leyton (Rutgers University and D.I.M.A.C.S.), A Generative Theory of Shape
7. M. Massironi, (Verona University), The Space of Representation and the Representation of Space
8. G. van Tonder (Kyoto Institute of Technology), Order and Complexity in Naturalistic Landscapes
9. D. Viswanath (UC Berkeley), Perceptual Representation of Surfaces and Objects and the Implications for Design
10. J. Willats (Birmingham University), Some Structural Equivalents Shared by Paul Klee's Paintings and Children's Drawings
11. A. Zimmer (Regensburg University), Visual Art and Visual Perception: An Uneasy Complementarity
12. S. Zucker (Yale University), Visual Computations and Visual Cortex

The conference will be organized by the Mitteleuropa Foundation, Bolzano, Italy ( <> ) Applications should be sent to Liliana Albertazzi (

Important Dates:
Deadline for abstract submissions: 15 April 2004
Conference: June 7-9, 2004
Dr Roberto Poli, PhD
Editor-in-chief of Axiomathes, Kluwer:
Papers and other information
preferred e-mail:
Dynamic ontology conference:

Monday, January 05, 2004

A Legal Mess

Much judicial commentary (and some professorial commentary) on the "doctrine of chances" fails to distinguish between
(i) the probability that a random selection of instances from some appropriate reference class will produce a conjunction of some specified states or values (e.g., "in instance 1 -- random draw number 1 --, event of type X occurs" and "in instance 2, event of type X [again] occurs" );

(ii) the probability, given a conjunction of of some specified states [such as in #(i)], that a criminal defendant caused those states of affairs.

Merely because the conjunction of events in situation #(i) above is highly improbable when instances of the reference class are chosen at random does not necessarily mean that some causal explanation -- such as "David Defendant caused such an improbable [i.e., rare] conjunction of events" -- is highly probable.

Why do legal professionals find it so hard to get a handle on the distinction between probabilities of type (i) and probabilities of type (ii)?

BTW: Does the following principle make intuitive sense to you?:

The occurrence of very improbable events and of very improbable combinations of events is highly probable.

Consider a restatement of this principle:

The occurrence of rare events and of rare sets of events is, over the long run [alternatively: "given a sufficiently large number of trials"], highly probable.

Dangerous Learning

A little statistical learning is a very dangerous thing.