Ambition, simplification and falsification

The questions, the deliverables, the methods, the projects, and the status of the research program.

As with most things, theories can be more or less ambitious, where the more ambitious theories have a greater risk of failure. We start with the idea of what it means to fail, and then go on to discuss what it means to be ambitious.

Both simplifications and falsifications are descriptions of how the model, the representation or the process, does not correspond to the 'correct' model. In both cases the 'correct' model typically is not known, but it is hoped that it exists and will be found by modifying the model appropriately. Typically, the simplification is incorrect by design, and the falsification is by inference from observations comparing the model to a 'gold standard' or to a population (Turing test, analysis of variance). Both are documented and important parts of the research project.

The primary focus for falsification are the following deliverables:

1. single skill
• the representation of a single skill, and its execution into action (without or with integrated perception)
• the comparison to observed behaviour depends on the execution of the skill into action, a model of the real-time consequences of the action in the physical world, and a model of the observations that can be made of this physical action. The comparison is therefore at a very indirect level.
• the sexual reproduction (meiosis) of a single innate skill, including natural variation and the possibility for mutation
• we are not yet able to map genetic information into attributes of innate skills. We are therefore dependent on the concept of competence, i.e. that the proposed mechanism could produce the representation of the skill -- even if it does not correspond to the actual mechanism. Falsification comes through the challenge to others to produce a more competent and more elegant mechanism.
• the innate skills and processes involved in generating a learned skill from mimicry
• we are not yet able to map brain-based skills, even though research in neuroscience is making progress. We are therefore dependent on the concept of competence, i.e. that the proposed mechanism could produce the representation of the skill -- even if it does not correspond to the actual mechanism. Falsification comes through the challenge to others to produce a more competent and more elegant mechanism.
• the innate skills and processes involved in generating a learned skill from discovery
• Falsification again depends on competence.
2. group-level skill-sets with coordinated skills
• the representation of each of the skills, their execution into action, and their information exchange for coordination
• The information exchange in in the physical world and therefore open to observation.
• the sexual reproduction (meiosis) of group-level skill-sets, including gender-based skills such as for mating
• The main challenge is competence, but it may be possible to predict 'errors' that should be observable.
3. a complete inventory of skills for a single individual (innate and learned)
• the storage requirements for the innate skills in DNA
• the storage requirements for both innate and learned skills in the brain
4. the evolution mechanisms for single skills and group-level skill-sets with coordinated skills
• innate skills
• learned skills passed from generation to generation (apprenticeship)

Falsification of the main hypotheses is indirect, through the deliverables above:

1. that skills, at both the individual and group level, can be investigated through their information representation & processing requirements - in DNA and in the brain,
• evaluated through the competence of the models in (A) above
2. that there is an 'inner language', like a programming language, that represents skills,
• Evaluated through the competence of the mechanism and its contribution to the compression of a complete skill set. Open to be challenged by 'better' mechanisms.
3. that there are multi-individual complementary skills that enhance the fitness of a group, such as mating, hunting in packs, or herd-based defense against predators,
• should be directly observable - biological data
4. that the complementary skills are coordinated by a skill-based information exchange (communication) between the individuals,
• should be directly observable - biological data
5. that the evolution of skills at both the individual and at the group level can be investigated via information representation and information processing mechanisms,
• depends on the evaluation of (A) to (D) above
6. that there is an evolution in the structure and functionality of the 'inner language' as well as in the information exchanges between coordinated skills
• Note: I speculate that Chomsky's universal grammar is a consequence of the evolution of the structure and function of this 'inner language'
• Note: I speculate that the variety of human languages is a consequence of the evolution in the structure and functionality of information exchanges
• evaluated through the competence of the language model to support increasing complexity in the skills
7. that learned skills can be passed from one generation to the next through mechanisms such as mimicry, i.e. that animal apprenticeship can function analogously to evolution,



A theory is more ambitious if it carries more information. Models can be part of feasibility studies, i.e. carry the assertion that if it were not to be falsified, it would propose to be a possible part of the mechanisms that would be sufficient to explain the target behaviour by simulating it. The target behaviour could be narrow or wide in scope.

Theories and model can be more ambitious by claiming to be competent to account for the target behaviour by modelling all aspects of the mechanisms that produce the behaviour. They can be even more ambitious by claiming to be complete in their accounts of the target behaviour.