Cultural knowledge evolution considers how agents can evolve their knowledge through communicating and adapting their knowledge. Although cultural evolution is often modelled by copying behaviour, knowledge may require alternating use (deduction) and induction.
Cultural evolution is the application of evolution theory to culture. It has been applied to various aspects of our life in societies: from customs to languages, from boat shapes to company structures [Messoudi, 2011]. It may be addressed computationally through multi-agent simulations in which a population of agents interacts through games that are played randomly [Steels, 2012]. It is possible to test hypotheses by precisely crafting the rules used by agents in games and observing the consequences.
One protocol that has been repetitively used for testing culture transmission is the whisper game: someone whispers a sentence to the ear of someone else who repeat it to someone else, etc. In the end it is observed the large variation between the initial message and the final one. This variation is one important feature of evolution theory: it allows to generate new variants that are selected. Research in cultural evolution posit that unfaithful transmission is the main source of variation. However, some voices discuss the necessity of noisy transmission for generating variations [Acerbi, 2020].
We study cultural knowledge evolution, i.e. we consider knowledge as the culture that agents evolve [Euzenat, 2017; Bourahla et al., 2021]. Knowledge is abstract: it is not something that can be found or observed from the environment. It can be transmitted directly, e.g. by teaching or by copying theories, but it can also be transmitted through observing how others use it. The difference in this case is that transmission works in two steps: the transmitter uses its knowledge to provide examples and the receiver observes the examples and induces knowledge again. Given enough examples, it may be considered that transmission would be faithful. However, very often agents learn from a small number of examples. Moreover, even with a large number of examples, there may be several representations (knowledge) of them.
These characteristics make that knowledge transmission may have different characteristics from the classical whisper game. Indeed, the whisper game is analogical: the message is both the transmission means and the object to be transmitted. In the case of knowledge, the examples are the transmission means and knowledge is the object to be transmitted. In the whisper game, the transmission is unreliable because of the transmission mechanisms; in the knowledge case, it is unreliable because of this intermediary induction step.
We would like to show that variation maybe found in the transmission process without resorting to imperfect and noisy communication. Because of this two-steps process, it is possible to design experiments using exact methods which generate variation. The goal of this topic is to do it in order to assess this possibility and its impact on variation.
The principle of such a game would be the following:
The purpose of this topic is to investigate the properties of such a non analogical transmission:
Finally, knowledge transmission may be studied within a population across generations. For instance, it is possible to develop the same kind of experiments, but with agents having an offspring proportional to the accuracy of their knowledge. The question here is whether this allows to improve knowledge or to counter drift from the initial knowledge (provided that it was mostly correct).
The work could be expected to unfold as follows:
References:
[Acerbi, 2020] Alberto Acerbi, Cultural evolution in the digital age, Oxford university press, Oxford (UK), 2020
[Bourahla et al., 2021] Yasser Bourahla, Manuel Atencia, Jérôme Euzenat, Knowledge improvement and diversity under interaction-driven adaptation of learned ontologies, Proc. 20th AAMAS, online, pp242-250, 2021 https://moex.inria.fr/files/papers/bourahla2021a.pdf
[Bourahla et al., 2022] Yasser Bourahla, Manuel Atencia, Jérôme Euzenat, Knowledge transmission and improvement across generations do not need strong selection, Proc. 21st ACM international conference on Autonomous Agents and Multi-Agent Systems (AAMAS), (Online), pp163-171, 2022 https://moex.inria.fr/files/papers/bourahla2022a.pdf
[Euzenat, 2017] Jérôme Euzenat, Communication-driven ontology alignment repair and expansion, in: Proc. 26th IJCAI, Melbourne (AU), pp185-191, 2017 https://moex.inria.fr/files/papers/euzenat2017a.pdf
[Mesoudi, 2011] Alex Mesoudi, Cultural evolution: how Darwinian theory can explain human culture and synthesize the social sciences, Chicago university press, Chicago (IL US), 2011 See also: Alex Mesoudi, Andrew Whiten, Kevin Laland, Towards a unfied science of cultural evolution, Behavioral and brain sciences 29(4):329–383, 2006 http://alexmesoudi.com/s/Mesoudi_Whiten_Laland_BBS_2006.pdf
[Steels, 2012] Luc Steels (ed.), Experiments in cultural language evolution, John Benjamins, Amsterdam (NL), 2012
Links:
Reference number: Proposal n°3069
Master profile: M2, research oriented.
Advisor: Jérôme Euzenat (Jerome:Euzenat#inria:fr).
Team: The work will be carried out in the mOeX team common to INRIA & Université Grenoble Alpes. mOeX is dedicated to study knowledge evolution through adaptation. It gather permanent researchers from the Exmo team which has taken an active part these past 15 years in the development of the semantic web and more specifically ontology matching.
Laboratory: LIG.
Place of work: The position is located at INRIA Montbonnot (near Grenoble, France) a main computer science research lab, in a stimulating research environment.
Perspectives: There is possibility to pursue in PhD, especially related to knowledge transmission.
Procedure: Contact us and provide vitæ and possibly motivation letter and references.