Evolution, generative entrenchment and the bounds of rationality

1
Evolution, generative entrenchment and the bounds
of rationality

• Konrad Talmont-Kaminski
• Marie Curie-Sklodowska University

2
Two issues

• Bounded rationality theory claims all reasoning
  heuristic in nature
• Why should this be the case?
• Humes problem of induction defines the field of
  possible epistemic processes, both for reasoning
  evolution
• What about development of new heuristics?
• Humes problem of induction forces development of
  new heuristics to proceed by broadly evolutionary
  means

3
Bounded rationality
4
Heuristics all the way up

• Herbert Simon The Sciences of theArtificial 3rd
  ed. 1996
• Perfect rationality a bad model
• Satisficing not optimising
• Heuristics
• Rules-of-thumb
• Heuristics all the way up
• Adaptations to scientific theories
• Simple heuristics used outside bounded
  rationality
• Kahneman Tversky
• Dual process accounts of reason

5
Fast, frugal, etc.

• Bill Wimsatt Re-engineering Philosophy of Limited
  Beings 2007
• Broad characterisation of heuristics
• Fallible
• Frugal (and fast, too)
• Systematically biased
• Problem transforming
• Have specific uses
• Developed from other heuristics (Exapted)
• Is it heuristics all the way up?

6
Hume heuristics
7
Exaptation heuristics

• Jerry-built products of evolution
• Evolutionary history
• Building a Ferrari from a Morris MM
• Developmental pathways
• Exaptation
• Using existing traits for new functions
• Feathers in dinosaurs and birds
• Human reason
• Typical product of evolution
• Kahneman Tversky studies
• Collection of heuristics
• Is human reason bounded because of evolution?

8
Dual process accounts

• Jonathan Evans others
• Heuristics
• System 1
• Evolutionarily old
• Logical Thinking
• System 2
• Evolutionarily new
• Problem
• People do use logic
• Human reasoning is bounded
• How does system 2 work?

9
Humean dualism?

• Humes 2 systems
• Habits heuristics
• Reasons System 2
• Hume the original dual process theorist?
• No
• Problem of induction
• 250 years of looking for solution
• Problem affects deductive reasoning, also
• System 2, either
• Runs into problem of induction
• Heuristics that use logical features of
  environment

10
A naturalist Hume

• A different view of the problem of induction
• Not a problem
• A basic epistemic limit
• Humes fact of reasoning
• Heuristics are the response
• Heuristics wherever Humes Problem
• Relevance to evolution?

11
Heuristics all the way down

• Evolution is back-ward looking
• Adaptations suit previous environments
• Not necessarily future ones
• Environmental changes may lead to extinction
• Arms races (cheetahs gazelles)
• Evolution short-sighted due to problem of
  induction
• Adaptations are also heuristics
• Humes problem is the fundamental epistemic limit
• Determines what evolutionary processes possible
• Determines what cognitive processes possible

12
Open-endedness andgenerative entrenchment
13
Evolution engineering

• Contrast between
• Evolutionary processes
• Small changes
• Every step must be satisficing
• Engineering projects
• Novel solutions
• Only end product satisfices
• Evolutionary landscape
• Wheel
• Evolutionary products more limited?
• In one sense, yes
• Most limits due to
• Evolutionary histories
• Developmental paths

14
Open-ended bounded rationality

• Human reason bounded but open-ended
• Develops and obtains new abilities
• Consists of a set of heuristics but
• Develops new heuristics
• Exapts existing heuristics to novel functions
• Open-endedness makes it possible to transcend
  particular limitations
• While still remaining bounded
• Open-endedness key trait of evolution cognition
• Engineered systems either
• Closed end-products and of little interest
• Open-ended

15
Step-wise development GE

• Humes problem forces
• Open-ended systems to develop in step-wise manner
• Generative entrenchment (GE)
• Existing heuristics make new heuristics possible
• Wheel required for automobile
• The more connections the more entrenched
• New transport systems existing infrastructure
• Human pyramid
• GE results from open-ended development
• History/development paths significant for all
  open-ended systems
• Development of open-ended engineered systems will
  have basic traits of evolutionary change

16
Evolution artificiality

• Artificial systems
• Etymological root - artifice
• Adapted to their environment
• By evolution
• By engineers
• Some traits explained in terms of environment
• Atrophy of eyes in cave animals
• Structure of the tire

17
Conclusions

• Humes problem of induction
• The basic epistemic limit
• Applies to all artificial systems
• Forces the use of heuristics
• Limits how open-ended artificial systems can
  develop
• Forces the use of evolutionary processes

18
One more pyramid

• konrad_at_talmont.com http//deisidaimon.wordpress
  .com