Bootstrapping Social Intelligence

• Summary
• Outcomes of the Project
• Stages and Milestones
• Design of the Web Sites
• Some Issues to be Determined by the Participants

Summary

A group of academics that work on various aspects of social intelligence will interact via the software on a sequence of web sites, upon the subject of social intelligence. The software will use evolutionary and other self-organising mechanisms to promote the development of knowledge beyond that of any of the individuals who participate. There will be an initial period where the participants will interact via a simple mailing list (or existing participatory software like Hypernews or Wiki), followed by three periods of interaction on each web-site. As well as informing the research of the participants in the field on social intelligence, the task of each period of interaction is to determine the design of the next web site. In this way it is hoped that the collective evolution of knowledge that occurs at one web site will can be utilised in structuring the next evolutionary process. Thus the sequence of web-site intermediaries implements a bootstrapping process, allowing the collective knowledge, the design of the web-sites and the participants' understanding to co-evolve.

Outcomes of the project

• The active debate, consensus building and other interaction that occurs at the sites should inform the individuals' understanding of social intelligence, as a result of the emergent content of the sites and from the experience of participating.
• The bootstrapping process should result in new insights into the design of participatory web-sites that build upon the inputs of and interactions between the participants. The end result is unpredictable, but may represent a new paradigm for such software or may be an innovative mix of existing participants. At the very least a lot will have been learnt about what works and what does not.
• The sites themselves, once the interaction has finished, will be browseable by the public, as permanent records of the collective knowledge gained during this project.

Stages and Milestones

Stage 1.

There will be an initial period of discussion between the group of researchers using existing mediation software (e.g. Wiki, Hypernews or a simple web-based mailing list). The primary objective of this stage will be the setting of objectives for the first web-site and specify the outlines of its design. This stage will last 6 months.

During this stage a workshop will be held to brainstorm ideas for the web-sites: its objectives, design and initial content. This workshop will have an important social function - that of enabling the participants to establish social contact. This is particularly important given that the majority of the interaction between them will be electronic.

Stage 2.

The first web-site will be specified in detail and implemented. The system will be tested by the core team by a discussion of the detailed specification. This stage should last about six months.

Stage 3

The participants discuss and interact starting with the subjects agreed upon during the first workshop. However it is anticipated that new subjects and classifications will quickly develop and take over as the focus of interaction. It is important that the participants have their own work in the broad area of social intelligence in order to ground their interactions, otherwise it could become too self-referent and closed.

A major task of the first period of interaction is to determine the design of the second web-site including the way the software structures the interaction, methods of evaluating the result and the initial content of the site.

Stage 4

The design indicate by the outcome of the present web-site is turned into a specification. This may require a certain degree of interpretation. The design is then implemented and tested. It is probably necessary that the programmer join in the interaction in the latter part of the interaction in order to 'firm up' the specification so that it can be implemented.

Stages 5 and 6

This is simply an iteration of stages 3 and 4 with a new web-site.

Stages 7 and 8

This is another iteration of stages 3 and 4 with another web-site.

Stage 9

A final workshop is held consisting of the participants and other interested parties in order to evaluate the project. Topics would include: the outcomes in terms of the resulting structures on the web-sites; the experience of participating and how it informed or influenced their thinking on social intelligence; the consequences of the experiment in terms of the design of participatory mediating software.

Design of the web-sites

The detailed design of the web-sites can not be specified n advantage, since the main point of this proposal is that the design of each web-site is determined by the preceding period of interaction using the previous web-site in a sequential 'bootstrapping' process.

However there are several main paradigms of self-organisation upon which the design of the first web-site might be based.

An associative network

An associative network is where the links between nodes are given some sort of weight or priority, and the strengths of these vary according to the number of times they are traversed (or associated). Classical rules for strength adjustment date from Hebb. An example of this mechanism has been demonstrated by Bollen and Heylighen at the Principia Cybernetica Project.

A shared semantic network

A network of nodes, with labelled links allows the construction of a semantic network. This can be collaboratively constructed in order to build up a shared mental map. The benefits and properties of a simple shared mental map have been investigated by ??. Inference on this network could be implemented by means of defeasible inheritance.

An evolutionary algorithm

An evolutionary algorithm allows for the creation and destruction of nodes and links according to their success. Variation in the nodes could come from versions of the node created by the participants, whilst nodes that are not linked to within a certain time could be downgraded or eliminated. Given these two operations the nodes at the site would undergo an evolutionary process.

Database update

Perhaps the weakest analogy that could be used is that of a database. All contributed nodes can be seen as records on a database. In this paradigm the main processing can be seen as a matter of producing the answers to queries but it is also possible that it could perform some consistency checks and either tag conflicting data or resolve the inconsistencies

A conference

The last structure is that of a simple conference or newsgroup. In this version the site merely provides a passive structure for the dialogue to proceed. It could provide facilities for the easy construction of sub-nodes, replies, links and indexing but would engage in no automatic processing of the information.

Some issues to be determined by the participants

There are also a number of specific issues which will have to be settled.

• Motivation to participate. The projects will fail if the participants do not interact sufficiently to enable significant emergent outcomes. It is probable that the desire for mutual success, general interest and social contact would not be sufficient on their own. Possible mechanisms could include something like a contracted amount of participation, but this is difficult to enforce and unlikely to encourage a high quality of input. A mechanism that assigns credit or otherwise publicises the effort put in might have more success.
• Confidentiality. Once the perio of interaction has finished the end product will be made accessible to the public, as a non-interactive web-site. However, there is the question of whether to allow the public to watch' the proceedings of the site as it develops. The pros for this are that an audience can improve the quality of the interaction and may motivate some of the participants; the cons are that this may inhibit participants from risking making 'mistakes' and thus stifle some of the innovation.
• Anonymity. Allowing contributions to be anonymous might encourage lively debate and promote a more collective approach, but it might also lessen motivation to participate and allow a greater level of acrimony. It would also make credit assignment difficult.
• Pseudonyms. Some studies suggest that being able to distinguish and identify sources by name is important for the emergence of social embeddedness. However the labels used do not have to be real names. With a small selection of researchers such pseudonyms will be easily guessed, but it is possible that each participant could choose a set of pseudonyms to use.
• Assignment of credit. For academics (as for others) it is important that they are publically assigned credit for the intellectual work they put in. If contributions are not anymous, this is a lessor problem, but still it may be that individual contributions are not preceived as sufficiently significant. Maybe some overall measure of the importance of contributions can be seperately and continually compiled - this could be based on some of the citation-based statistics.
• Duration of collective memory. Death or forgetting is an important mechanism in evolution. It allows for greater adaptability because resources are re-distributed to newer trial solutions. It thus reduces the chance that the population may get 'locked-in' to unproductive avenues. On the other hand t is necesssary that the essence of the better ideas be kept in some form to be reused in subsequent development.
• Controlling participant membership. Ideally there the same group of academics should be the participants throughout the project, since their remembered experience of their interactions is an important factor in the evolution of the knowledge. However, in reality, it is likely that some participants wil be unable to continue for the full generation and will need to be replaced. Public participation is not planned due to the potential problems this might bring. Thus there will need to be some mechanism for the suggestion of new members and the agreement of participants to the new entrants. Possibly it should be the leaving participant to nominate some 'suessors', which then would be ratified by the other participants.
• Maintaining relevance to social intelligence. It is inevitable that the subject matter of the interactions will develop in unforseen directions. In this lies the point of the project - to foster the construction of knowledge that would otherwise not have occurred. It is also possible that the subject matter would 'wander' far from the field of social intelligence. It is possible that some 'soft' feedback mechanism as to the relevance of the interaction content might be included, such as its connectivity to key 'root' nodes.
• Limiting severe structural change. The inclusion of the ability to alter the structure of the interactions as a result of the interactions is a potentially powerful mechanism for evolution. However, structural change that is too rapid can disrupt and even completely halt the evolutionary process. For this reason it is suggested that the mechanisms of the first web-site be unchangeable, that the first interaction should suggest changes in these for the second site and the ability to effect the mechanisms dynamically only be allowed in the third site, if at all.
• Linking into and from other web-sites. As well as being grounded in the knowledge and developing research of the participants, the content of the sites will also relate to the wider body of knowledge, as is represented by the content of the world-wide web. It is expected that interactions will include links out to relevant sites elsewhere, but to be a full part of the web it will need to be able to be linked to as well. The possible confidentiality of the sites will make this difficult.
• Maintaining the context of interactions. In normal conversation and academic writing there is a rich context for any communication. This context gives the meaning to the words and sentences. In a large and dynamic web-site, it may be easy to lose sight of the context, which might then lead to a high level of misunderstanding and loss of intended meaning. Some mechanisms that help to 'frame' the interactions in terms of its role within the history of the interaction and the location in the site are needed.
• The provision of global search tools. Once these sites get large it will become very difficult for any single individual to keep track of the whole. This is not necessarily a bad thing as it encourages participants to specialize in particular areas of the site. Global search tools such as search engines for the site enable one to make syntactic searches on the content, and thus effect the way that the participants interact.
• Security measures. Adequate but unobtrusive security measures will need to be put into place, especially if the site is viewable by the public.
• The accessible information horizon. As noted in the sections on the context of interaction and the provision of global search tools, the locality (or conversely globally) of the 'view' within the site effects its dynamics. A more local approach would tend to promote variety and parallelism while a global approach promotes coherence and consistency.
• Automatic inference or induction. In addition to the contributions of the participants there is potential for automatic additions and alterations, particularly in the structure and strength of links between nodes. This could be based on a number of different paradigms, including: Hebbian strengthening of associations, the inference of relevant new links and evolutionary selection of nodes based on browsing or linking statistics.
• Governance of the project. The nature of the project means that its exact course will not be forseeable so that there might inevitably arise key decisions about the shape and structure of the project. There hs to be some procedure for making these, even when there is not a consensus on these by the participants. Some formal democratic mechnism is suggested, especially fr the initial stages of the project.
• Automatic notification. Once a site becomes large and complex it becomes onerous for an individual particiapant to keep up with all the developments therein. This can slow the pace of interaction because reactive dialogue is difficult. This is not necessarily a disadvantaage, as this would promote more consideed responses and encorage specialisation into certain areas of the total site. However it can also discourage contributions and lose the dynamic flavour of the interations. If one is automatically notified of contributions that relate to one's own contributions thi can keep the iterest level up.