New evidence for systematicity in infants’ curiosity-driven learning.

Abstract

Decades of research demonstrate that infants’ learning is sensitive to task features. However, what level of complexity best supports learning is unclear. Moreover, infancy studies work typically employ carefully-designed experiments with complexity determined a priori. Whether infants systematically generate a particular level of difficulty during everyday, curiosity-driven exploration is therefore unknown. Twomey & Westermann’s (2017) model of visual curiosity-driven learning predicted that infants will generate intermediate task complexity (cf. Kidd, Piantadosi & Aslin, 2012). Experiment 1 tests this hypothesis, while Experiment 2 extends this work to a naturalistic environment. In Experiment 1, we developed a new shape priming paradigm. Complexity was instantiated as perceptual (color, shape) differences between a continuum of 2D exemplars. Infants were primed with peripheral exemplars, and their exploration of the remainder of the category was recorded using a screen-based eyetracker. Infants systematically directed first looks to intermediate complexity stimuli, irrespective of prime, and subsequently generated intermediate difficulty exploratory sequences. Cluster analysis demonstrated that infants exhibited one of two exploratory styles, generating either high-intermediate low-intermediate complexity sequences. Experiment 2 extending our shape priming paradigm to a naturalistic environment using 3D-printed stimuli and head-mounted eyetracking. Stimulus edges differed in a continuum from corners to rounded. Preliminary data show a contrasting pattern of exploration to the 2D environment: after corner primes, 18-month-olds selected exemplars of greatest complexity, while after rounded primes, the infants preferred the most similar exemplars. Together, these studies offer new evidence that infants as young as 12 months actively impose structure on their learning environment.

Full text

New evidence for systematicity in infants’ curiosity-driven learning
Han Ke1*, Gert Westermann1 Ben Malem2 & Katherine Twomey3*
1Lancaster University 2King’s College London, 3University of Manchester
*h.ke@lancaster.ac.uk, katherine.twomey@manchester.ac.uk
Literature Review Experiment 2
Experiment 1
Conclusions and future directions
! "#$#%&'()*#+,-$.&%.#).(%.)/-0%-.$1)#234,&%./,-)/$)$#-$/./5#).,)0#%.6&#$),0).(#)
#-5/&,-+#-.7)$6'()%$).(#)',+34#2/.8),0),9:#'.)0#%.6&#$);<,6-=#&)>)?#%&/-=7)
@AABC)D,5%'EFG#$( >)H%E#$7)IJJKLM
!N-0%-'8)$.6*/#$).83/'%448)#+34,8)'%&#06448F*#$/=-#*)#23#&/+#-.$)O/.()
',+34#2/.8)*#.#&+/-#*)a"prioriM)P(#.(#&)/-0%-.$)$8$.#+%./'%448)=#-#&%.#)%)
3%&./'64%&)4#5#4),0)*/00/'64.8)*6&/-=)#5#&8*%87)'6&/,$/.8F*&/5#-)#234,&%./,-)/$)
.(#&#0,&#)6-E-,O-M)
! QO,+#8)>)P#$.#&+%--1$);IJ@KL)+,*#4),0)5/$6%4)'6&/,$/.8F*&/5#-)4#%&-/-=)
3&#*/'.#*).(%.)/-0%-.$)O/44)=#-#&%.#)/-.#&+#*/%.#).%$E)',+34#2/.8);'0M)D/**7)
R/%-.%*,$/)>)S$4/-7)IJ@ILM)T23#&/+#-.)@).#$.$).(/$)(83,.(#$/$7)O(/4#)
T23#&/+#-.)I)#2.#-*$).(/$)O,&E).,)%)-%.6&%4/$./')#-5/&,-+#-.M)
Shape&Priming&paradigm.&12-month-old&infants (label&condition:&n&=&20,&no-label&
condition,&n&=&20)
-Shape&Priming&paradigm&with 3D&printed&
objects&-12-month-old&infants (N =&18)
-6&prime-test&pairs
-Head-mounted&eye&tracking
First-pick&analysis
Distance
from prime
Prime
122
210
323
433
Touch-sequence cluster analysis
This work was supported by the ESRC International
Centre for Language and Communicative Development
(LuCiD), an ESRC Future Research Leaders fellowship to
KT and a British Academy/Leverhulme Trust Senior
Researcher Fellowship to GW. The support of the
Economic and Social Research Council [ES/L008955/1;
ES/N01703X/1] is gratefully acknowledged.
S=%/-7)0/&$.)3/'E)$(,O#*)%)$8$.#+%./')3%..#&-M) UI);VL)W)IJMA@7)3)W)
MJJX
QO,+#87)DM)TM7)>)P#$.#&+%--7)YM);IJ@KLM)Z6&/,$/.8F9%$#*)4#%&-/-=)/-)
/-0%-.$[)S)-#6&,',+36.%./,-%4)%33&,%'(M)Developmental"Science7)
;H'.,9#&)IJ@\L7)@]@VM)(..3[^^*,/M,&=^@JM@@@@^*#$'M@I\IA
D/**7)ZM7)R/%-.%*,$/7)_M)QM7)>)S$4/-7)"M)`M);IJ@ILM)Q(#)Y,4*/4,'E$)#00#'.[)
a6+%-)/-0%-.$)%44,'%.#)%..#-./,-).,)5/$6%4)$#b6#-'#$).(%.)%&#)
-#/.(#&).,,)$/+34#)-,&).,,)',+34#2M)RG,_)H`T7)K;cL7)@]BM)
(..3[^^*,/M,&=^@JM@VK@^:,6&-%4M3,-#MJJV\VAA
R&/+#[),-#)#2.&#+#)$(%3#7)@c$
Q#$.[)&#+%/-/-=)0,6&),9:#'.$7)VJ$
Look-sequence&cluster&analysis
Type&1 Type&2
First-look&analysis
<#$)0/&$.)4,,E$)%&#)-,.)&%-*,+48)
*/$.&/96.#*d)
eI;VL)W)@JMVJ7)3)W)MJ@\
N-0%-.$1)0/&$.)4,,E$)%&#).,)4,O),&)
/-.#&+#*/%.#)*/$.%-'#)$./+64/
!N-0%-.$)0&,+)@I)+,-.($),4*)$(,O)$8$.#+%./')3%..#&-$,0
#234,&%./,-M
! a,O#5#&7)4,,E/-=)%-*)3/'E/-=)#234,&%./,-)$(,O$)*/00#&#-.)
$8$.#+%./')3%..#&-$ ] O/44)/-0%-.$)$(,O)$/+/4%&)3%..#&-$),0)
4,,E/-=)/-)T23#&/+#-.)I)%$)/-)T23#&/+#-.)@f
!T234,&%.,&8 $.84#$)*,)-,.)*/00#&)9#.O##-)3%&./'/3%-.$7)$,)
O(%.)*&/5#$).(/$)$8$.#+%./'/.8f)R&/+#f)_#b6#-'#)4#-=.(f)
g/$.%-'#f)
Distance
from prime
First-look
Silent Label
10.32 0.35
20.38 0.32
30.14 0.16
40.14 0.16
Type&1 Type&2 Type&3
N-0%-.$1)0/&$. 3/'E %&#).,)(/=(#$. */$.%-'#)$./+64/M
Acknowledgement
Head-mounted&eye&tracking&video&