GROUP 07 - Conference notes - Part 1

This is a summary of the key findings and themes from the conference with some annotations from me as to their relevance to Cnx. Some have obvious and direct implications for us - others are more amorphous and have implications for where we go in the future. As mentioned this is Part 1, to be followed by Part 2 of key findings and then detailed notes on some interesting papers and workshops.

Key Findings:

• Do not design only to provide a solution – design for seduction as well.

• Designing for strangers vs. designing for people that know each other through offline channels

• Need for information gatekeepers – expertise concierge

• Content seeding for UGC – critical to get going

• Wikis more useful for transactive memory sharing when knowledge is distributed amongst members of group

• Risk of social capital – this was a big theme – defines how much people participate and how they present themselves online.

• Wikipedia - .1% editors contribute half the value – keep your top producers happy and provide them incentives. Elite editors contribute most value and this domination is increasing. Keep Elite editors happy.

• Tagging was another big theme – usefulness and combination of tags with other systems.

  • Users are influenced by tags they see in the community. Seeing good tags leads to creating good tags

  • High quality tags are applied more often

  • Tags can have personal value (memory aids – list-making) vs. social value (convey added meaning)

  • How can a tagging system show users all the tags they want to see considering the limited screen real-estate? Show “high-quality” tags first.

  • Evaluating tagging behavior – generating metrics and design heuristics. Based on Tag growth, Re-use, Non-obviousness, Discrimination Frequency and temporal patterns.

    • Social system should encourage tag re-use by suggesting tags others have already applied in the system.

    • High tag growth and low reuse implies personal bookmarking

    • Any discernible trends in behavior? eg. tagging increases near exam time – provide a system that can support this behavior.

  • Comparison of tagging vocabularies across services – tag suggestions have profound effects on tagging vocabularies

  • Normalization at search time can lead to good results (normalization across tag variants, polysemy, synonymy, scope)

• Reusing informal information from discussions etc – periodic information distillation by SME editors.

• Collaborative generation of content benefits from incremental formalization of concepts and thoughts. Provide tools to do so.

• Improve performance of recommender systems by using social network information eg. create recommendation based NOT just on the most similar neighbors but also based on things used/created by friends and co-clique members

• Expertise-locater systems – this was another major theme. As technology facilitates increased collaboration – the need to find information on short notice increases. Find – Reach – Network. Almost like LinkedIn but for expertise location. This can be an interesting feature we can offer in the future. Find the person with the required expertise – Reach through your personal network.

  • Associations only go from topics to people

  • Populated not by static user created profiles but by their content – this can be key for us. We know what type of content authors create in our system. We can auto-populate an expertise database with this information

  • Expertise in social context – a range of expertise – not just THE expert.

• Group formation was another key theme

  • Metrics for evaluation of group formation – formation, performace, percieved satisfaction

  • Collaborative activities with dynamic feedback which guides reflection and improves

  • Feedback can be expert evaluation, peer evaluation and automated communication analysis (word count, word pattern etc.)

  • How to paticipate in a group so everybody likes me vs. how to participate so the group is more productive.