This week our task was to explore each of our respective ideas for apps. My idea was generated by trip to the Grace Hopper Conference listening to a lecture by one Jennie Lees, who was talking about the inadequacy of video game suggestion software. She discussed how the algorithms that help suggest games to play use stereotypes and poorly made assumptions about the kinds of games that different kinds of people like to play. I looked into some of the resources out there and discovered three major sources of video game suggestions:
Online quizzes:
These range in length and complexity, but all seem to be pretty terrible. They lump users into categories with questions such as "Do you like killing things?" and "Do you have a short attention span?" They use aspects of games to find out what game would be a good fit. In other words, if you do not already know what kind of games you like, you are unlikely to find a good one here. They also use a very limited source of games and do not include indie games. In addition, these quizzes are made by online quiz sites rather than a site dedicated to video games. Here and here are two examples.
Forums:
Another pretty solid source of video game suggestions are online forums like reddit, which has a thread solely for video game suggestions, as well as the forum page of video game sites themselves. You must be pretty deep into the video game community to be able to find these threads. However, the information provided is generally pretty good as it is crowd-sourced, heuristic knowledge.
Suggestion software:
Finally, there are sites dedicated solely to suggestions. Game Finder is a site that has a directory of games with human-generated similar games. The main downside to this site is the limited base of games. However, once you select a game that IS featured there is a wide variety of games, not featured in the directory, that are suggested. This is probably the best option out there, but again, all the featured games are reviewed by people and thus it does not take advantage of the benefits of big data.
Another suggestion site is TasteKid, which is a suggestion site for movies, books, music, authors, shows, and games. It is definitely a more generic version of Game Finder and does not provide the detail, but its UI is more streamlined. I haven't of course been able to actually see the algorithms that go into suggesting the games, but from the little exploring I did it seems that the suggestions are pretty generic.
For both of the above suggestion sites, you MUST know about games that you have already played. Furthermore, if you potentially would like a broad variety of games rather than just more of one type of game, this is not taken into account.
Conclusions:
It seems to me that some aspects of suggestion software that would be good to incorporate into a new app would be using information that is not game related, taking more time to get to know the user, and combining human-generated content with big data. This would of course need to be balanced with the aspects that are already being done - using information from previously played games, suggesting games quickly, and having an established infrastructure for the kind of content being displayed (forum vs. simple display).
Datasets:
The next step would be identifying sources of data in order to form an ontology. We discussed in our meeting using the Steam API, which looks very easy to use. It returns data in JSON, VDF, and/or XML (yay!). Based on the wiki, it appears that we can retrieve a lot of information about a user if we have their Steam ID which we can obtain securely by using the OpenID framework: users click a button, are redirected to Steam where they securely log in, and then we obtain their verified ID. From my previous research, it does not appear that any other software sites are utilizing this function.
Next steps:
My next steps are to contact Jennie Lees and ask her for direction and thoughts on the subject. I did approach her after her lecture about this idea and she seemed amenable to working with me through email. Next, I would need to grab some test data from the Steam API and see what kinds of things I can use from there. I am looking forward to presenting this information to the group!
Online quizzes:
These range in length and complexity, but all seem to be pretty terrible. They lump users into categories with questions such as "Do you like killing things?" and "Do you have a short attention span?" They use aspects of games to find out what game would be a good fit. In other words, if you do not already know what kind of games you like, you are unlikely to find a good one here. They also use a very limited source of games and do not include indie games. In addition, these quizzes are made by online quiz sites rather than a site dedicated to video games. Here and here are two examples.
Forums:
Another pretty solid source of video game suggestions are online forums like reddit, which has a thread solely for video game suggestions, as well as the forum page of video game sites themselves. You must be pretty deep into the video game community to be able to find these threads. However, the information provided is generally pretty good as it is crowd-sourced, heuristic knowledge.
Suggestion software:
Finally, there are sites dedicated solely to suggestions. Game Finder is a site that has a directory of games with human-generated similar games. The main downside to this site is the limited base of games. However, once you select a game that IS featured there is a wide variety of games, not featured in the directory, that are suggested. This is probably the best option out there, but again, all the featured games are reviewed by people and thus it does not take advantage of the benefits of big data.
Another suggestion site is TasteKid, which is a suggestion site for movies, books, music, authors, shows, and games. It is definitely a more generic version of Game Finder and does not provide the detail, but its UI is more streamlined. I haven't of course been able to actually see the algorithms that go into suggesting the games, but from the little exploring I did it seems that the suggestions are pretty generic.
For both of the above suggestion sites, you MUST know about games that you have already played. Furthermore, if you potentially would like a broad variety of games rather than just more of one type of game, this is not taken into account.
Conclusions:
It seems to me that some aspects of suggestion software that would be good to incorporate into a new app would be using information that is not game related, taking more time to get to know the user, and combining human-generated content with big data. This would of course need to be balanced with the aspects that are already being done - using information from previously played games, suggesting games quickly, and having an established infrastructure for the kind of content being displayed (forum vs. simple display).
Datasets:
The next step would be identifying sources of data in order to form an ontology. We discussed in our meeting using the Steam API, which looks very easy to use. It returns data in JSON, VDF, and/or XML (yay!). Based on the wiki, it appears that we can retrieve a lot of information about a user if we have their Steam ID which we can obtain securely by using the OpenID framework: users click a button, are redirected to Steam where they securely log in, and then we obtain their verified ID. From my previous research, it does not appear that any other software sites are utilizing this function.
Next steps:
My next steps are to contact Jennie Lees and ask her for direction and thoughts on the subject. I did approach her after her lecture about this idea and she seemed amenable to working with me through email. Next, I would need to grab some test data from the Steam API and see what kinds of things I can use from there. I am looking forward to presenting this information to the group!