5 replies [Last post]
Joined: 09/24/2012
Groups: Go Science

In order to answer this research question, an experiment was made in 2 phases.

1) A live experiment with the creation of a "Virtual Soloist Groups" (rules here: https://fold.it/portal/node/993167#comment-27535)

Results:
Formed 04/18/2014, deleted 05/05/2014 (on request of several top players because this resulted in a perturbation of the "real" group rankings).
Group ranks after this period: prediction 12, design 13, overall 18.
Best ranks: 6th in puzzles 880 & 882.

2) What would have happened if a Soloist Virtual Group had been present on the latest 25 puzzles?

Method:
Taking the score of the best Soloist without group in 25 latest puzzles
Taking the corresponding rank and score from the group ranking
retaining the best 20 scores
Multiplying by 2 (Overall scores are based on 40 best out of 50)

Result:
1318 points
Overall rank 7
Best rank: 3 on puzzles 946 and 962.

Discussion
There is a positive bias compared to the first live experiment (where only volunteers soloists participated). It's possible that, in the time of the live experiment, players participating were more productive due to competition spirit (this would make a negative bias). A positive bias comes from the fact that many beginners might not want to join a group: "Virtual Soloist Group" would have relatively less experimented players. This has to be verified.

On a sample puzzle (961), there was 104 soloists players on a total of 188 'active players' (= who got positive score improvement): 55%. That means that the probability to win for a Soloist Virtual Group is much higher than for each individual groups. Number of active players of the top groups is given here:
http://fold.it/portal/node/998128

On sample puzzle 961, current group ranked 7 in the overall list (Gargleblasters) and in this puzzle had only 7 active players (on a total of 22 regular players for this group). They did as good as the 104 virtual members of a "Soloist Virtual group".

Conclusion

Group production seems much more competitive than a sum of individual competitors.

(The homo oeconomicus of mainstraim neoclassical economic theory might not be as competitive as the theory supposes... Actually, selfish players have few to gain in groups, most of the commons being on the open access Wiky and public recipes. This suggests that working for a "common group gain" - and related group members recognition- is a high incentive for top Foldit players).

Joined: 09/24/2012
Groups: Go Science
On soloist ranking

Are group members favored in the soloist ranking ?. If this is true, the above conclusion should be revised.

First, it's technically possible to cheat as suggested in my comment here: https://fold.it/portal/node/1998890#comment-29394
and proved by truestone here:
https://fold.it/portal/recipe/45935

Note that cheating is technically possible within a group, but a pure soloist can also cheat coming into an open group, copying solutions or recipes, going out. Than means that we never have the certitude that a soloist is a "pure soloist", neither that a group member is a potential cheater or not.

Second, even if the community and group rules are fully respected, a group member could be influenced by a top shared solution in order to find an idea for his/her poor current soloist solution. If not copy-pasting structures or loading a group solution as guide.
Moreover, group members share expertise and advices in chat and group forum. Mainly, they share "group recipes". Group recipes are like devprev. If there is a new good idea, the group has always an advantage on the all Foldit community (including other groups).

Conclusion

-Group members are favored in the soloist ranking

Possible solutions

1-
Individual profiles could show group history
; with this info, we could identify "pure" soloists and make their "pure" ranking. There are many problems with this solution (see discussion referred in the hyperlink).

2-johnmitch just created a group named solo. In this group, no share are admitted. The further advantage is that being in this group does not easily allow you to go out, joining an open group, copying, coming back. The condition is that this group should be closed or at least moderated. Being in this group is a guaranty that you play the solo competition with full honesty. this "group" would be the reference for a pure solo competition for the ones who like it.

3-An irreversible status in player's profile could be "Pure soloist" => "Potential group member"., Beginners start with being "pure soloists". Potential group members are players that ever went to any group. This is irreversible. It is then possible to make an informative subranking of pure soloists. This might be an incentive for beginners (advanced players cannot compete in this ranking any more). We could even imagine that advanced players are automatically excluded from this subranking. Why? Because they participated in global chats and they got info that beginners cannot have. This "pure soloists" competition would then also be limited to the first year or say, first 6 months, or first 100 puzzles.

Joined: 08/09/2010
groups == learn

I have one thing clear: if I had not joined a group I would not be a good soloist.
The main incentive for a beginner to join a group is how much he will learn from veterans.
And with these tests we should know whether the soloist have used recipes of their old groups ;)

Joined: 09/24/2012
Groups: Go Science
Indeed

You learn a lot by
-doing yourself
-reading the wiki, looking at the videos
-getting tips from the peer (global chat, veteran chat, group chat)

The "small group" give you the following extra:

1) Confinent communications in adapted language or culture

2) Motivation:
-groups are more "human" in many ways than an anonymous crowd of players
-group recognition: you receive a lot of "thanks" when you help the group in a way or another; and you are admirative or grateful when you see a good result from a peer
-group competition: a common goal for group rank in competition with other groups
-group feeling (being proud to be part of a group that found a famous ebola solution :)
-"fill the gap" feeling: trying to find a place to be useful depending on other group members skills and attitudes
-reciprocity feeling (having received from mentors, you feel the responsibility to render to the group, or to reciprocate to beginners = chain reciprocity)

3) High level expertise building
-Learning from working on real top solutions (not only pictures like in the wiki, but tridimensional touchable solutions)
-In many groups, the players communicate the way they got to a solution (succession of recipes used etc), this helps to build expertise
-I learned scripting from a real programmer (Jean-Bob) in my native language; hand folding tips from Ch Gamier, and some real science by a PhD biochemist (roukess)
-I wonder "how they do" in top groups (like AD); this makes me think and chalenge my (also social) creativity

4) Diversity, buffer "countries":
-you can visit a group, not an individual player, come with something from your "country" and go back with some interesting ideas
-the semi-open social structure makes diversity and performance

Conclusion
The social structure of Foldit (combining various levels of communicating buffers) is imitating real life efficient structures

Joined: 04/24/2014
Groups: None
This work

This is pretty interesting to read - thanks for doing the work on this. :)

DataGuy's picture
User offline. Last seen 4 years 36 weeks ago. Offline
Joined: 01/27/2014
Groups: None
Interesting read.

Interesting read.

Sitemap

Developed by: UW Center for Game Science, UW Institute for Protein Design, Northeastern University, Vanderbilt University Meiler Lab, UC Davis
Supported by: DARPA, NSF, NIH, HHMI, Amazon, Microsoft, Adobe, RosettaCommons