CES talk
Simplicity and informativeness in the cultural evolution of language Jon W. Carr, Kenny Smith, Jennifer Culbertson, Simon Kirby Centre for Language Evolution School of Philosophy, Psychology and Language Sciences University of Edinburgh
Pressures shaping language
Language
Pressures shaping language
e l p m Si Learning
Language
Pressures shaping language
e l p m Si Learning
Language In
e v i t a form
Communication
Kinship terms are simple and informative English
Northern Paiute
4
Informative
3
2
1
0 Kemp & Regier (2012)
0
Simple
50
100
Kinship terms are simple and informative English
Northern Paiute
4
Informative
3
2
1
0 Kemp & Regier (2012)
0
Simple
50
100
Lab experiments
Learning-only
Lab experiments
Learning-only
Communication-only
Lab experiments
Learning-only
Learning + Communication
Kinship terms are simple and informative English
Northern Paiute
4
Informative
3
2
1
0 Kemp & Regier (2012)
0
Simple
50
100
Learning and communication pressures 4
3
Informative
2
1
0 0
Simple
50
100
Learning and communication pressures 4 Learning-only
3
Informative
2
1
0 0
Simple
50
100
Learning and communication pressures 4 Learning-only
3
Informative
2
1 Communication-only
0 0
Simple
50
100
Learning and communication pressures 4 Learning-only
3
Informative
2
1 Learning + communication
Communication-only
0 0
Simple
50
100
Learning and communication pressures 4 tuge tuge tuge
tuge tuge tuge
tuge tuge tuge
tupim miniku tupin
tupim miniku tupin
tupim miniku tupin
poi poi poi
poi poi poi
poi poi poi
Learning-only
3
Kirby, Cornish, & Smith (2008)
mega
egewawa
megawawa
egewuwu
megawuwu
ege
wulagi
1 Learning + communication
Communication-only
0 gamenewawu
gamenewawa
gamenewuwu
nemone
piga
kawake
kapa
gakho
wuwele
nepi
newhomo
kamone
gaku
hokako
2
Informative
egewawu
pihino
gamene
Kirby, Tamariz, Cornish, & Smith (2015)
0
Simple
50
100
Kirby, Tamariz, Cornish, & Smith (2015)
Iterated learning gives rise to informative spatial terms
Carstensen, Xu, Smith, Regier (2015)
Iterated learning gives rise to informative spatial terms
Carstensen, Xu, Smith, Regier (2015)
Experiment 1
Angle
Stimuli
Size
Angle
Stimuli
Size
Angle
Stimuli
Size
Angle
Stimuli
Size
Conditions Angle-only
Size-only
Easy to learn but low informativeness
Angle & Size
Informative but hard to learn
Training phase
Test phase
Results Angle-only
Results Size-only
Results Angle & Size
Result: Learnability advantage for the less informative systems
Experiment 2
Iterated learning
Results 1 category
2 categories
3 categories
4 categories
Simplicity
Simplicity
Informativeness
Simplicity
Informativeness
Simplicity
Informativeness
Measuring simplicity
Measuring simplicity
Measuring simplicity
4×8
4×8
Measuring simplicity
1×1 3×3
1×1
3×4
4×8 1×1 1×8 1×2 2×6 4×8 1×4
1×2 1×6
2×3
What makes language informative? Property
Example
Expressivity A system of many categories is more informative than a system of few categories
>
Convexity A system of convex categories is more informative than a system of nonconvex categories
>
Learning
Communication
Pressure for simplicity
Pressure for informativeness
What makes language informative? Property
Example
Expressivity A system of many categories is more informative than a system of few categories
>
Convexity A system of convex categories is more informative than a system of nonconvex categories
>
Learning
Communication
Pressure for simplicity
Pressure for informativeness
What makes language informative? Property
Example
Expressivity A system of many categories is more informative than a system of few categories
>
Convexity A system of convex categories is more informative than a system of nonconvex categories
>
Learning
Communication
Pressure for simplicity
Pressure for informativeness
What makes language informative? Property
Example
Expressivity A system of many categories is more informative than a system of few categories
>
Convexity A system of convex categories is more informative than a system of nonconvex categories
>
Learning
Communication
Pressure for simplicity
Pressure for informativeness
What makes language informative? Property
Example
Expressivity A system of many categories is more informative than a system of few categories
>
Convexity A system of convex categories is more informative than a system of nonconvex categories
>
Learning
Communication
Pressure for simplicity
Pressure for informativeness
What makes language informative? Property
Example
Expressivity A system of many categories is more informative than a system of few categories
>
Convexity A system of convex categories is more informative than a system of nonconvex categories
>
Learning
Communication
Pressure for simplicity
Pressure for informativeness
What makes language informative? Property
Example
Expressivity A system of many categories is more informative than a system of few categories
>
Convexity A system of convex categories is more informative than a system of nonconvex categories
>
Learning
Communication
Pressure for simplicity
Pressure for informativeness
Our experiments
What makes language informative? Property
Example
Expressivity A system of many categories is more informative than a system of few categories
>
Convexity A system of convex categories is more informative than a system of nonconvex categories
>
Learning
Communication
Pressure for simplicity
Pressure for informativeness
Carstensen et al.
Conclusions The pressure from learning yields simple languages, but this can manifest itself through multiple semantic properties: Expressivity Loss of words/categories to aid learning Convexity Reorganization of the meaning space to aid learning The pressure from communication has the same effect on convexity as learning does, making it difficult to identify the causal mechanism. In iterated learning, the pressure for simplicity favours semantic category systems that are convex – informativeness comes along for the ride.
Vielen Dank!
Pressures shaping language
Language
Pressures shaping language
e l p m Si Learning
Language
Pressures shaping language
e l p m Si Learning
Language In
e v i t a form
Communication
Kinship terms are simple and informative English
Northern Paiute
4
Informative
3
2
1
0 Kemp & Regier (2012)
0
Simple
50
100
Kinship terms are simple and informative English
Northern Paiute
4
Informative
3
2
1
0 Kemp & Regier (2012)
0
Simple
50
100
Lab experiments
Learning-only
Lab experiments
Learning-only
Communication-only
Lab experiments
Learning-only
Learning + Communication
Kinship terms are simple and informative English
Northern Paiute
4
Informative
3
2
1
0 Kemp & Regier (2012)
0
Simple
50
100
Learning and communication pressures 4
3
Informative
2
1
0 0
Simple
50
100
Learning and communication pressures 4 Learning-only
3
Informative
2
1
0 0
Simple
50
100
Learning and communication pressures 4 Learning-only
3
Informative
2
1 Communication-only
0 0
Simple
50
100
Learning and communication pressures 4 Learning-only
3
Informative
2
1 Learning + communication
Communication-only
0 0
Simple
50
100
Learning and communication pressures 4 tuge tuge tuge
tuge tuge tuge
tuge tuge tuge
tupim miniku tupin
tupim miniku tupin
tupim miniku tupin
poi poi poi
poi poi poi
poi poi poi
Learning-only
3
Kirby, Cornish, & Smith (2008)
mega
egewawa
megawawa
egewuwu
megawuwu
ege
wulagi
1 Learning + communication
Communication-only
0 gamenewawu
gamenewawa
gamenewuwu
nemone
piga
kawake
kapa
gakho
wuwele
nepi
newhomo
kamone
gaku
hokako
2
Informative
egewawu
pihino
gamene
Kirby, Tamariz, Cornish, & Smith (2015)
0
Simple
50
100
Kirby, Tamariz, Cornish, & Smith (2015)
Iterated learning gives rise to informative spatial terms
Carstensen, Xu, Smith, Regier (2015)
Iterated learning gives rise to informative spatial terms
Carstensen, Xu, Smith, Regier (2015)
Experiment 1
Angle
Stimuli
Size
Angle
Stimuli
Size
Angle
Stimuli
Size
Angle
Stimuli
Size
Conditions Angle-only
Size-only
Easy to learn but low informativeness
Angle & Size
Informative but hard to learn
Training phase
Test phase
Results Angle-only
Results Size-only
Results Angle & Size
Result: Learnability advantage for the less informative systems
Experiment 2
Iterated learning
Results 1 category
2 categories
3 categories
4 categories
Simplicity
Simplicity
Informativeness
Simplicity
Informativeness
Simplicity
Informativeness
Measuring simplicity
Measuring simplicity
Measuring simplicity
4×8
4×8
Measuring simplicity
1×1 3×3
1×1
3×4
4×8 1×1 1×8 1×2 2×6 4×8 1×4
1×2 1×6
2×3
What makes language informative? Property
Example
Expressivity A system of many categories is more informative than a system of few categories
>
Convexity A system of convex categories is more informative than a system of nonconvex categories
>
Learning
Communication
Pressure for simplicity
Pressure for informativeness
What makes language informative? Property
Example
Expressivity A system of many categories is more informative than a system of few categories
>
Convexity A system of convex categories is more informative than a system of nonconvex categories
>
Learning
Communication
Pressure for simplicity
Pressure for informativeness
What makes language informative? Property
Example
Expressivity A system of many categories is more informative than a system of few categories
>
Convexity A system of convex categories is more informative than a system of nonconvex categories
>
Learning
Communication
Pressure for simplicity
Pressure for informativeness
What makes language informative? Property
Example
Expressivity A system of many categories is more informative than a system of few categories
>
Convexity A system of convex categories is more informative than a system of nonconvex categories
>
Learning
Communication
Pressure for simplicity
Pressure for informativeness
What makes language informative? Property
Example
Expressivity A system of many categories is more informative than a system of few categories
>
Convexity A system of convex categories is more informative than a system of nonconvex categories
>
Learning
Communication
Pressure for simplicity
Pressure for informativeness
What makes language informative? Property
Example
Expressivity A system of many categories is more informative than a system of few categories
>
Convexity A system of convex categories is more informative than a system of nonconvex categories
>
Learning
Communication
Pressure for simplicity
Pressure for informativeness
What makes language informative? Property
Example
Expressivity A system of many categories is more informative than a system of few categories
>
Convexity A system of convex categories is more informative than a system of nonconvex categories
>
Learning
Communication
Pressure for simplicity
Pressure for informativeness
Our experiments
What makes language informative? Property
Example
Expressivity A system of many categories is more informative than a system of few categories
>
Convexity A system of convex categories is more informative than a system of nonconvex categories
>
Learning
Communication
Pressure for simplicity
Pressure for informativeness
Carstensen et al.
Conclusions The pressure from learning yields simple languages, but this can manifest itself through multiple semantic properties: Expressivity Loss of words/categories to aid learning Convexity Reorganization of the meaning space to aid learning The pressure from communication has the same effect on convexity as learning does, making it difficult to identify the causal mechanism. In iterated learning, the pressure for simplicity favours semantic category systems that are convex – informativeness comes along for the ride.
Vielen Dank!
