To investigate the foundations of situated knowledge evolution we need an approach that:
- is general enough so that results can encompass various cases;
- is flexible enough so that many specific settings may be established;
- provides an explicit representation of knowledge in order to communicate it to others;
- allows for continuous local adaptation depending on the situation;
- allows for both theoretical and experimental work.
No single approach of the state of the art offers all these features.
Thus, mOeX will develop the unique combination of knowledge representation and experimental cultural evolution methods.
Knowledge representation provides formal models of knowledge;
experimental cultural evolution provides a well-defined framework for studying situated evolution.
We do not intend to replace symbolic representation, but to complement it.
The reasons why these approaches are well adapted are the following:
- Agents usually cannot wait for slow processes to terminate: they will apply adaptation operators allowing them to communicate;
- Agents need an explicit representation of knowledge in order to communicate it to others;
- Agents do not need that all knowledge is correct before acting;
- Agents do not have a global view of other agents' knowledge: they need a distributed solution with partial knowledge.
Our methodology involves the following three tasks interacting together in a constant feedback:
- Designing games and knowledge adaptation operators;
- Performing experiments for testing potential properties of such operators;
- Analytically proving such properties or their conditions.
Thus, mOeX adopts a dual theoretical and experimental strategy:
determining which adaptation operators provide expected knowledge properties and designing actual mechanisms such that results can be assessed experimentally.
Both approaches cross-fertilise: theory by suggesting adequate operators and properties to test and experiments by providing results to explain.
Methodological approach.
Finally, in order to ensure the repeatability and reusability of experiments we aim at developing a software platform to support this approach.
It is, in particular based on formally describing experiments so that they can be repeated and searched [Euzenat 2022a].
The multi-agent cultural evolution software that we use is available at https://gitlab.inria.fr/moex/lazylav and experiment are stored in our repository https://sake.re.
We are repurposing our experiment reports as Jupyter notebooks.