New and Emergent World models Through Individual, Evolutionary, and Social Learning

Main page

Last update: 12-01-2005
Last changes: Call for papers for AISB Emergent Artificial Societies Symposium, PowerPoint project summary added
Latest news:
and registered, this page relinked

Project duration: 01-09-2004 - 31-08-2007
Project participants: see
consortium page
Project funding:
NEW TIES is a research project funded by the European Commission via the 6th Framework's Future and Emerging Technologies open scheme.

The original aim as formulated in personal discussions is to create (initiate + have evolve/emerge) an artificial society with high mental and linguistic capabilities.In particular, it would be great to see whether/when/how they start wondering about the origins of their world and themselves. Pushing it to the extreme: see if they ask themselves if WE exist. Name it artificial or evolutionary philosophy? (cf. artificial life, evolutionary economy, etc). Interestingly, these ideas were partly inspired by a piece of literature: Stanislaw Lem's short story called "Non Serviam" (originally in S. Lem, A Perfect Vacuum, English translation published by Harcourt Brace Jovanovich, 1983, also included in the famous AI collection Douglas R. Hofstadter and Daniel C. Dennett (Editors), The Mind's I: Fantasies and Reflections on Self & Soul, Basic Books Inc. 1981, pp. 296-317).

MAIN OBJECTIVES (from the project proposal)
The goal of the project is to realize an evolving artificial society capable of exploring its virtual world and developing its own view of that world. The long-term target is to learn how to design agents that are able to adapt autonomously to, and then operate effectively in, environments whose features are not known in advance. The main pillars of the envisioned research are world models and the learning mechanisms generating these. We shall not implement specific training facilities or feedback systems rewarding the learning of world models for its own sake. Instead, we are interested in emerging world models powered by a basic survival game. In order to obtain the expected emergent features, we shall work on a very large scale compared to that of today’s common practice with respect to the size of the agent population and agent complexity. Defining culture as knowledge structures shared among agents that reflect aspects of the environment, including the other agents, our first scientific objective is:

1. To develop an artificial society with an emergent culture.
This objective can be broken into the following sub-objectives:
1.1. To identify which system components carry the knowledge structures that make up world models.
1.2. To develop mechanisms to monitor the development of world models.
The population's abilities to develop advanced skills bottom-up consist of individual learning, evolutionary learning, and social learning. This motivates the second main objective.

2. To realise a powerful “emergence engine” consisting of a well-balanced combination of individual learning, evolutionary learning, and social learning.
This objective can be broken into the following sub-objectives:
2.1. To develop an understanding of the mutual effects of the three types of learning on each other and on the development of the individuals and the whole population.
2.2. To develop mechanisms for adjusting the balance between the three types of learning.

We consider individual learning and evolutionary learning as the two basic mechanisms and expect that social learning has a (re)combinative effect on the learning processes significantly boosting the population’s adaptive capabilities and enabling the society to rapidly develop an "understanding" of the world collectively. One of the main innovations of this project lies in the way we approach social learning. It will be implemented in such a way that it allows passing knowledge explicitly (i.e., not only by imitation, but via direct messages) to others within the same generation. The corresponding scientific objective is:

3. To develop, evaluate, and use a range of social learning mechanisms that allow sharing knowledge with other members of the population.
Our approach to this objective is twofold. On the one hand we will study various protocols of information dissemination, implying the following sub-objective:
3.1. To set up communication mechanisms and investigate their effects on the efficiency of social learning mechanisms.
On the other hand we need to address issues about the form and content of messages. This leads to the second sub-objective, which relates to (the emergence of) language and its use:
3.2. To implement a framework enabling the emergence of the communication and cooperation essential to social learning.

PROJECT SUMMARY (Abstract of the proposal)
The project is concerned with emergence and complexity in socially-inspired artificial systems. We will study large systems consisting of an environment and an inhabitant population. The main goal of the project is to realize an evolving artificial society capable of exploring the environment and developing its own image of this environment and the society through cooperation and interaction. We will work with virtual grid worlds and will set up environments that are sufficiently complex and demanding  that communication and cooperation are necessary to adapt to the given tasks. The population's weaponry to develop advanced skills bottom-up consists of individual learning, evolutionary learning, and social learning. One of the main innovations of this project is social learning interpreted as passing knowledge explicitly via a language to others in the same generation. This has a synergetic effect on the learning processes and enables the society to rapidly develop an "understanding" of the world collectively. If the learning process stabilises, the collective must have formed an appropriate world map. Then we will probe the collective mind to learn how the agents perceive the environment, including themselves, and what skills and procedures they have developed to adapt successfully. This could yield new knowledge and surprising perspectives about the environment and the survival task. The project represents a significant scale-up beyond the state-of-the-art in two dimensions: the inner complexity of inhabitants and the size of the population. To achieve and explore highly complex organisms and behaviours, very large populations will be studied. This will make the system at the macro level complex enough to allow significant behaviours (cultures etc) to emerge in separate parts of the system and to interact. To enable this we will set up a large distributed computing infrastructure, and a shared platform to allow very large scale experiments in a p2p fashion. .