Representing and reasoning about artificial kinds: On the sources of relevant information, and the r

1
Representing and reasoning about artificial
kindsOn the sources of relevant information,
and the relevance of sources of information

• Tamsin C. German
• Department of Psychology
• University of California, Santa Barbara

2
Function represented directly via a core
knowledge system?
Keil (1992, 1994, 1995) Functional-teleological
mode of construal perhaps having arisen to
handle instances of design in biology. Pinker
(2002) Intuitive engineering deals with our
ability to make and use tools, perhaps shared
with primates (and crows?).
3
Artifact concepts relate information across core
domains
Object mechanics
4
But we dont consider the goals of all social
agents to be equal
And we have numerous highly specified objects
with restricted functions
5
What information forms the core of the concept?
DESIGN
SHARED CONVENTION
6
1. That core (whatever information it is based
on) appears to play a role in function based
problem solving
Function demonstrated
Control
7
Simple model of conceptual structure underlying
functional fixedness
8
(Defeyter German, 2003, Cognition German
Defeyter, 2000, Psychonomic Bulletin Review)
Median time (in seconds) to use target object,
by age and condition
9
(No Transcript)
10
2. That core (whatever information it is based
on) plays a role in function assignment and
naming
DESIGN vs USE INVENTOR Made for trapping
bugs. OWNER Used it to collect raindrops. What
is it really for? Why?
ORIGINAL vs CURRENT NAME INVENTOR Called it
a TOG. OWNER Calls it a FEP What is it
really ? Why?
11
German Johnson, 2002, Journal of Cognition and
Development
12
Investigating what information is relevant to the
function
Used for
Made for
Carrying bottles
Catching fish

• Function judgment Whats it for? Catching fish
  or carrying bottles?
• Category judgment What is it? A fish catcher or
  a bottle carrier?

13
Idiosyncratic single user current function
(Defeyter German Hearing, in press)
14
Conventional everybody current function
(Defeyter German Hearing, in press)
15
So a distinction between information about
shared conventions in categorization versus in
function judgment?
Places constraints on possible functions
Contain soup? ? Stub out cigarettes? ? Cover
head from rain? ? Protect head from axe? ?
But the same mechanical structure does not place
any constraints on possible morphological forms
that could be the name of the category the
relationship between names and structure is
arbitrary.
16
New study (in development) to test this idea

• Fast mapping of word meanings (Markson Bloom,
  1997).
• Rapid learning of artifact functions (Casler
  Kelemen, 2005)
• Earlier, not necessarily restricted to design
  function.
• Sought to investigate possible differences in
  fate of discarded information in learning names
  versus functions.
• Prediction Correct names retained and mistakes
  rapidly discarded. Correct functions and
  (plausible) mistaken functions retained.

17
So sources of relevant information versus
relevance of sources of information?

• Temptation to deal with evidence from (more
  recent) problem solving tasks as somehow
  different with explanations requiring
  external factors (education, age,
  executive function etc).
• Note reverse chronology in current presentation
• Performance demands/models are no less important
  in understanding categorization/judgment/word
  learning tasks as they are to problem solving.
• Concepts are not just for thinking/talking (e.g.
  Hood, Carey Prasada, 2000).

18
Coda priming of non design/conventional
functions?
?
?
19
Thanks to

• Danielle Truxaw, Max Krasnow, Chantelle Woods.
• Members of the Cognition and Development Lab,
  UCSB
• Margaret Defeyter, Jeanette Ingwerson, James
  Hatton and Jill Hearing, all at the University of
  Northumbria, UK.
• ESRC
• The British Academy