Three Puzzles + Lecture Notes on Features

From: HARNAD Stevan (harnad@coglit.soton.ac.uk)
Date: Sat Feb 28 1998 - 20:08:23 GMT


Hi All:

Here are the lecture notes on Features. They will also be on the Web,
in the class's Skywriting Archive.

Come prepared to discuss (and solve, if you can, and if it's possible)
the three puzzles below.

Also come prepared to talk about the Schyns, Goldstone & Thibaut
article and commentary.

Cheers, Stevan

Perception, Categorisation and Features

Does the world start for us as a "blooming, buzzing confusion"?

Natural gaps in the confusion

"Prepared" feature-detectors (e.g., frog bug-detector and human facial
expression detectors)

The "classical" theory of categorisation:

We recognise KINDs of things; we know what kind they are (i.e.,
which category they belong to) on the basis of FEATURES.

Features can be PARTS (like legs or corners) or they can be
"DIMENSIONS" like size or colour, or they can be COMBINATIONS of
dimensions or parts, like shapes or melodies or or textures.

So according to the classical theory of categorisation, we recognise
kinds of things by "detecting" a feature or a set of features. The
features can be SIMPLE, like a colour or a leg or a sound, or they can
be COMPLEX combinations, like four legs and a tail, or round, green and
striped, or high, shrill and intermittent.

To recognise something as being of certain kind, or belonging to a
certain category, the thing must (1) HAVE features on the basis of
which we can tell it apart from other kinds of things, and (2) we must
be able to DETECT and USE those features.

Just Noticeable Differences (JNDs).

Instruments and extensions of our senses.

Some puzzles:

Puzzle 1:

Watanabe's Ugly Duckling Theorem:

All things share an infinite number of features.
So two things cannot be of the same kind because they share more
features than they do with things of a different kind: All things share
the same number of features.

So we have to select some features and ignore others. HOW DO WE KNOW
WHICH FEATURES TO SELECT?

Positive and Negative Instances.

Selective attention and abstraction.

Memory Limitations.

[Remember Funes the Memorious.]

Concrete Categories
[the kinds of things we can see or hear directly with our senses]

Abstract Categories

Individuals: Jack, my bed, the Murray Building

Kinds: Person, bed, Building

Higher-level kinds: Organisms, Furniture, Real Estate

Properties: moves, flat, square-shaped

"Completely Abstract" kinds and properties?

Goodness, Truth, Beauty

Puzzle 2:

What are the features of "games" (or goodness, or truth, or beauty)?

IF WE DON'T KNOW THE FEATURES, OR THERE ARE NO FEATURES, HOW DO WE KNOW
WHAT GAMES AND GOODNESS AND TRUTH ARE (AND AREN'T)? How do we manage to
categorise them anyway?

Prototypes and typicality.

Some things look like better or more typical examples of their category
than others: Robins are better examples of birds than penguins of
ostriches. So is category membership a matter of degree? Is a robin more
of a bird than a penguin is?

Puzzle 3:

Objective and Subjective Categories.

Concrete categories are the ones we recognise directly with our senses:
How about our mental states? How do we recognise our moods?

Feeling angry, frightened, excited, depressed?

What are their positive and negative instances?
What are their features?

What about feeling AWAKE (not wide awake vs. barely awake, but AWAKE as
opposed to ASLEEP? What about feeling ALIVE (as opposed to DEAD)? or
CONSCIOUS (as opposed to UNCONSCIOUS)?

Do we really have a category of "feeling awake" "feeling alive" or
"feeling conscious"?

What is wrong here?



This archive was generated by hypermail 2b30 : Tue Feb 13 2001 - 16:23:20 GMT