The ACBE project goals are to:
The first work line will develop general models of artificial knowledge and belief evolution, involving games and agents that take part to cultural evolution experiments. An initial model will be provided to support the implementation of the simulator (WL2) and experiments (WL4). It will be constantly updated upon experiments and will reach a second version.
There is currently no such general model of cultural evolution games. Some attempts have recently been made albeit focussing on one single game. One important challenge of this work is that the designed models should be sufficiently precise so that (a) they can be processed by the simulator, and (b) they can be theoretically exploited in order to prove games or setting properties. Such models will offer the capability for agents to play several games and for games to be played by several types of agents.
An integrated simulator will be developed based on the models developed in WL1. It will enable to identify key parameters of cultural evolution and to straightforwardly experiment in a parametric and extensible way. The first iteration will process games with agents as defined in WL1. The second iteration of the simulator will support agents playing several games at once and different types of agents playing. This will also allow for specific experiment scheduling, such as synchronous playing, which is often used in some types of simulations.
The ability for the same agent to enter different games using the same kind of knowledge and beliefs is a point that will require a specific attention and will constitute a significant advance on the state of the art. The fact that different agents cannot enter the same set of games does not raise problem and will offer the opportunity for new experiments.
In order to support open and reproducible experimental science, this work line will provide a way to describe experiments, their hypotheses and their results so that they can be easily repurposed and rerun. Such precise descriptions are key to the independent reproduction of experiments. In a second time, these descriptions will be extended and exploited in order to be able to retrieve, select and compare a large base of experiments and their results.
The difficulty of this work is to provide formal descriptions that allow experiments and global analyses to be processed. A balance has to be found between the depth of descriptions and the capability to address a wide range of experiments.
The aim of this work line is to demonstrate that the promoted approach can be used to obtain significant results in experimental artificial cultural evolution. Based upon the simulator provided in WL2, cultural knowledge and belief evolution experiments will be carried out. This will also offer a first library of games and agents to be reused in further experiments. A first round of experiment will aim at reproducing existing experiments within the framework. This should provide a fast feed-back on its implementation. The second round of experiments will more specifically show the benefits of clearly identifying evolution mechanisms within agent models. Finally, the third experiments will provide evidence of the capacity of the system to support the integration of different kinds of agents simultaneously playing different games.
The goal of this work line is not simply to test the proposed environment: it aims at providing valuable research results in cultural knowledge and belief evolution, and specifically to its core with respect to evolutionary mechanisms. With respect to this, beyond the reports, deliverables are expected to be experiment records including experiment descriptions, results and analysis, and that will be published in papers.
This work line deals with the dissemination of the outputs of ACBE. It will promote the proposed approach and platform through open workshops and tutorials aiming to disseminate towards different audiences, with a specific goal to reach social sciences and the humanities, and opens to different culture types beyond knowledge and belief representation.
The dissemination task is of specific importance for ACBE as outreach to at least a part of researchers involved in cultural evolution research belongs to our objectives. For that purpose, paper publication about our progress will also be part of it.