au
The Social Route To Abstraction
Kristian Tylén , Riccardo Fusaroli , Pernille Smith , Jakob Arnoldi
The Experiment
Through multiple trials, individuals and dyads were presented with images
of aliens and asked to decide the appropriate course of action. They recieved feedback after each trial.
Figure 1: Example of a trial. The participants (individuals or pairs) had to decide
on a course of action using the number
keys 1 - 4:
=2
§: The Interacting Minds Center,
Aarhus University,
Jens Chr. Skous Vej 4,
8000 Aarhus,
Denmark
=3
=4
Ignore the peaceful, non-nutri-
tious aliens
Tap a valuable jelly from peaceful,
nutritious aliens
Kill dangerous aliens
Kill and tap jelly from dangerous, nutritious aliens
¶: School of Culture and Society,
Aarhus University,
Jens Chr. Skous Vej 4,
8000 Aarhus,
Denmark
Figure 2: Examples of stimulus items
from A: the training phase, and B: the test
phase. The descriminating features were:
1) legs (long or short),
2) arms (upwards or downwards),
3) body (with or without dots),
4) eyes (on stalks or not), and
5) color (green or blue)
A: Examples of training items
BA
A
0.8
0.70.8
0.7
0.7
0.6
0.6
Learning curves: Dyads present higher initial performance, intercept: β = 0.65, SE = 0.3, t = 2.15, p
= 0.03, and a steeper linear increase, β = 31.54, SE
= 10.33, t = 3.05, p = 0.002.
Furthermore, the slope of learning curves significantly predicts test performance, irrespective of condition, β = 2.36, SE = 1.06, t = 2.22, p = 0.026, see
figure 3C.
1
2
Session
A
∞: School of Business and Social Sciences,
Aarhus University
Bartholins Allé 10,
8000 Aarhus,
Denmark
0.6
0.5
3
BB
1 1
2 2
3 3
Session
Session
Session
0.7
1
Individuals
Dyads
0.6
0.6
Training: Dyads were more accurate than individuals, β = 0.32, SE = 0.04,
t = 7.97, p < .001.
A significant negative interaction between condition and complexity, β =
-0.13, SE = 0.05, t = -2.89, p=0.004, is due to the fact that dyads’ advantage
over individuals becomes smaller in the third session compared to the first,
see figure 3A.
0.7
0.5
0.7
Results
Individuals
Individuals
Dyads
Dyads
0.5
0.8
B: Examples of test items
B
0.6
0.5
There were three sessions with increasing levels of complexity.
In the low-level complexity condition, a single or combination of two distinct features would constitute a category. In the highest complexity condition combinations of more than two features out of a repertoire of five would
constitute a category.
£: The School of Communication and Culture,
Aarhus University
Jens Chr. Skous Vej 2,
8000 Aarhus,
Denmark
A
Correct
We hypothesize interacting dyads to have an advantage compared to individuals, especially when presented to problems of increasing complexity, because ...
i) cognitive diversity leads to broader search and more abstract represen[1 - 3]
tations
.
[2, 4]
ii) social interactions work against the problem of functional fixedness
.
The stimulus aliens varied in terms
of five binary features.
After 96 trials of training followed
a test with 8 aliens differing slightly in appearence (different colors
and shapes), but sharing the same
basic features.
Test: We observe a significant interaction between
condition and complexity, β = 0.33, SE = 0.16, t =
2.01, p = 0.045 indicating that dyads’ advantage is
bigger in the third session, when compared to the
first session, see figure 3B.
Correct
Correct
Stimulus
Are our capacities for abstract thinking and category-formation critically
facilitated by social interaction?
=1
§,∞
Correct
Introduction
§,∞
0.5
0.5
1
2
Session
Figure 3: Performance (proportion of accurate responses) by
individual and dyads over the three sessions (complexity levels).
A: performance in training phases,
B: performance in test phases,
C: Relation between learning curves during training and performance on test across conditions and sessions
1
3
C
2
3
Session
Session
9
8
Performanceon
on test
Performance
test
AARHUS UNIVERSITET
§,¶
Correct
§,£
7
6
5
4
5
6
7
Learningcurve
curve estimates
Learning
estimates
8
Discussion
Compared to individuals, dyads showed an overall performance advantage
and steeper learning curves. Furthermore, dyads were better than individuals at applying acquired knowledge to new settings especially at higher levels of complexity.
Our results suggest a link between dialogical interaction and cognitive
processes of abstraction. We speculate that cognitive processes of abstraction are facilitated through the explicit negotiation of complementary and/
[1, 2]
or conflicting perspectives
. Such circumstances create a pressure for
[3]
more abstract and flexible verbal representations .
References
1. Page, S. E. (2008). The difference: How the power of diversity creates better groups, firms, schools, and societies. Princeton University Press
2. Tylén, K., Fusaroli, R., Bjørndahl, J. S., Rączaszek-Leonardi, J., Østergaard, S., & Stjernfelt, F. (2014). Diagrammatic Reasoning: abstrac-
tion, interaction, and insight. Pragmatics and Cognition, 22(2), 264 - 283
3. Voiklis, J., & Corter, J. E. (2012). Conventional wisdom: negotiating conventions of reference enhances category learning. Cognitive
Science, 36(4), 607-634
4. Duncker, K. (1945). The Structure and Dynamics of Problem-Solving Processes. Psychological Monographs, 58(5), 1-112
2
Sess