Language Evolution and Computation Course at KIT Graduate School of Language Technology, Helsinki, 2003

Overview:

What is language, how did it come to be the way it is, and why are we humans unique in our possession of it? These are some of the most fundamental questions in linguistics and cognitive science. They go to the heart of what it means to be human. Over the last couple of decades there has been a remarkable renaissance in the field of Language Evolution, which aims to answer these questions. Researchers from a huge range of backgrounds, from Archeology to Neuroscience, are actively seeking to understand where language came from.

One approach to answering these questions that has been particularly popular in recent years is computational modelling. Bringing together techniques from machine learning, evolutionary computation, and multi-agent modelling, this novel methodology allows us to probe different theories of language evolution to see if they really work. Computational simulations allow us to run experiments on artificial populations to try and recreate some features of language evolution in silico.

This course has a dual purpose. Firstly, it will provide a review of the core concepts and latest findings in language evolution research. Secondly, it will explore how we can begin to use computational simulations to understand these findings. Any student taking this course will go away not only with a critical understanding of modelling techniques but also the foundation upon which these simulations are grounded: the facts of the evolution of human language.

Slides (Kirby):

• Lecture 1: Introduction to learning
• Lecture 2: Learning syntax
• Lecture 3: Evolving syntax
• Lecture 4: Learning bias and linguistic structure
• Lecture 5: Iterated learning
• Practical: The Talking Heads
• Final slides
  

Essay topics:

• close, detailed argumentation for a clearly expressed proposition
• accurate representation of relevant facts
• sustained relevance to the specific topic of the essay

1. Defend, or attack, the proposition that the grammatical structures of human languages are merely encodings of pre-existing conceptual structures which were not originally used for communication.
   
   
2. What is problematic, in evolutionary terms, with the idea that human language evolved to enable communication? 
   
   
3. What does it mean to claim that some aspect of language structure is 'self-organizing'? 
   
   
4. Are the basic units of human speech vowels and consonants? In what sense (e.g. psychological/articulatory/acoustic/computational) could this be true, and if it is true, how did this situation arise? 
   
   
5. How have computational simulations been used to explain the critical period for language acquisition?
   
   
6. Human languages seem to share certain universal properties (and certain logically possible types of language are very rare, or non-existent). These facts have been taken to imply the existence of an innate Universal Grammar. How have neural network simulations been used to argue against this, and how successful have these arguments been?
   
   
7. What are the similarities and differences between cultural evolution through iterated learning, and biological evolution through natural selection? How have both been used to explain the origins of language?
   

Background reading:

• Kirby, S. (2002). Natural Language from Artificial Life. Artificial Life, 8(2):185--215. [Download]

• Hurford, J. (2003). The Language Mosaic and its Evolution. In M.H. Christiansen and S. Kirby, editors, Language Evolution. Oxford University Press. [Download]
  
  
• Christiansen, M. and Kirby, S. eds (2003). Language Evolution. Oxford University Press. [Link]
  

Useful links:

• Language Evolution and Computation Bibliography
  
  
• Language Evolution and Computation Research Unit, Edinburgh
  
  
• Msc/PhD programme in the Evolution of Language and Cognition
  
  
• Talking Heads simulation software
  
  
• TLearn neural network simulation tool (see also Elman et al. (1996) Rethinking Innateness and Plunkett & Elman (1997) Exercises in Rethinking Innateness).
  

Lecturers:

• Simon Kirby [email: simon@ling.ed.ac.uk]
  
  
• Jim Hurford [email: jim@ling.ed.ac.uk]