Find me a problem solving corpus. I'm willing to pay $100 for each one you find

[urlwolf]

Are you good at finding information? This could be an interesting challenge. It could also be a very fast way of making $100 (in DC credits or paypal).

As you may know, I'm a cognitive scientist and I use corpora (large amounts of experience) to train models.
It is not very easy to find those, but given the huge amounts of information that is online, you may be able to find one for me! I need you to find large corpora of people doing ANY activity that has to do with problem solving. NOTE: the definition of problem solving is any activity where an agent (i.e., a human) needs to do a series of actions to reach a desired state. Do a search on what is considered problem solving in cognitive science (see http://en.wikipedia....wiki/Problem_solving). I'm more interested in the European tradition mentioned there.

What I need is a corpus of 'log files'. E.g., if you know a gameplaying community who has logs of their members playing since a long time ago, you probably have a winner.

Here is a description of some examples of the corpora that I would like to find:
http://www.andrew.cm...esada/corpusBasedPS/

These are videogames, a simulator, etc. But you can find me any activity where we keep logs for every participant; be creative. Another example would be the more 'classical' games, such as chess. A large database of chess games would be interesting too, but less so.

There's a list of features you need to fulfill for a corpus to be deemed as 'good'.

These are the features I need for a corpus to be 'valid':
1- The task can be captured in log files (e.g., a list of states and actions from the time the person starts playing -solving the problem- till the time the problem is solved, or the person has depleted her resources: e.g., running out of gas, bullets, crashing a plane, or losing all enablers so reaching the goal is impossible.
2- The logs are public, or the copyright owner of these logs doesn't mind us to use them for research
3- The task is complex and dynamic. Note: these semantically rich domains. If you find a large corpus of people playing tic tac toe, sorry, that is not what I need. I need realistic, complex activities
4- Large datasets (I.e., overall it adds up to several years of practice)
5- Each individual has many repetitions on the task under very simple conditions (e.g., landings on the same airport; games on the same 'level', etc)
6- In this corpus, you can find individuals who have been playing for a long while (e.g., weeks, or better yet, months).
7- The authors of the problem solving logs are approachable, and willing to participate in experiments doing more playing/problem solving

Since there are lots of videogames out there and people spend many hours playing, there must be plenty of opportunities to find corpora with these characteristics.

If you find me one or more of these large corpora, I'll pay you $100 each.
If you have any additional questions, please PM me.

Thanks

[Cpilot]

Hehe,
You oughta take a shot at Second Life.
If you can get log files from there it'll keep you busy for years. They have different ways to access the Grid maybe you can find something of interest there.

[urlwolf]

Thanks Cpilot,
I did take a look at second life.
It doesn't meet requeriments 5 and probably 7.
It's too under-constrained; I mean, it's a life! not a closed task like e.g., landing a plane. There's nothing that tells you when the task is 'done'. Or how good a solution is.

Keep them coming!

[KenR]

Wow. I didn't know you did that work. I am/was a clinical neuropsychologist and did some cognitive psychology as well. I did not do a lot of research in cognitive, but it is an interesting and challenging area.

Ken

[cthorpe]

What about http://www.mabiweb.com/index.php

It is a play-by-web site for a number of strategy board games.  If you select "Completed Games" on the left under "Game Manager" you can get a list of recently completed games.  When you select one, you can see the game board at the end of the game, and get a full log of all the moves made by the players.  The completed games list only has the last 7 days, but you may be able to get access to older logs if you ask.


Carl

[Laughing Man]

Haha seems I find a neat community with other psychologists. I'm a student in psychology right now. But my intent is to go into graduate school in cognitive science or industrial organization. So far most of my experience with cognitive psychology has been the lab I work in for Human-Computer interaction. I'd love to get an idea about what graduate schools have a strong program for cognitive psychology though. It's getting about that time where I start applying. 

[urlwolf]

Thanks cthorpe, good catch!
The problem I have with the games I can see at mabiweb is that they are two-person games; that's an entirely different ballpark. The whole point of my approach is that you know the environment -even though it's dynamic-. Having another player as an environment is overcomplicating things. Not to say that it's not interesting; there's plenty of good work on multiplayer games from within game theory. However, these are ultra-simple games.

I have had some other people contributing logs from shooter games. I have to think more about those.
Here's an example of a catch:

1)
The logging has already been made and is still going on. What we are talking about would be
- 2GB of compressed data
- over 4 years of collecting data
- over 5 million players total logged
- 1000+ still active

points of interest could be-
improvemt on map x
improvement on team red/blue
playing time / month
kills / time
deaths / time


2) according to the owner it should be used under Creative Commons
(Attribution, Noncommercial, ShareAlike) license
3) This is a shooter game with many different aspects, so rather complex
4) we are talking about several THOUSAND years of playing time
5) Definitely many repetitions, there are only so many maps
6) in some cases individuals have been playing for years
7) the authors are approachable, willing to do more logging,
There are still a few thousand active players, but getting them to participate would be your job, but that shouldn't be too hard.

This is kind of on the big side of things. If this is played with teams, it may be too complex. Ideally, it'd be one guy against an environment. The logs must contain a detailed description of what was on the screen when the person made a particular action. So I'd need both actions and states, that is only action logs or only states logs are not enough.


@Ken, Interesting. Have you stopped practicing because of your health? If so, I'm sure there's still lots you can do with your current knowledge that can be extremely useful just using the web. Plenty of interesting data!

@Laughing Man:
Maybe I can help you selecting a grad school. Shoot me a PM with the unis you have selected, and I'll do my best to do an impartial review.

[urlwolf]

So it looks like this is actually a quite difficult thing to find...
Nobody has claimed the $100. I posted this same task at rentAcoder and elance, an noone proposed anything remotely interesting...

[urlwolf]

Actually, this might just that I need to find the right forum for this to work.

Any suggestions? NOTE: newer games, like first-person shooters, are probably too complicated, since they are multplayer and people are teaming up and shouting what to do to each other when they are in the same room (my model would be deaf, so it's in a disadvantage here).

Simpler, older games (even strategy games) may be better.

[tinjaw]

urlwolf,

I say this in all honesty. I have talked to you before about doing a serious game project together. Let's do one that gets you the data you need. Sure, you won't be able to use it for your current project, but if there is a need for this type of data out there, lets generate it. PM me. Let's do this.