So I was playing around in Google Reader with the recently established stream feed on melative, when I saw some comments flowing in on usagi’s lively Cross Game post. ghostlightning mentioned my name, and it grabbed my attention. He touched on the relation of behavior and decisions which occur, leading to a newly found anime endeavor; given, it could/should be generalized for just about every type of media.
Rather than deeply ponder the thoughts, I’ll just start by having a representation of my own order of influence when it comes to taking on a series I may have never given time.
First, the mentioned sources:
- a - Direct/RL Recommendation
- b - [Professional] Internet Review
- c - Blogger Review
- d - Blogger Season Preview/Hype/Pimp
- e - Blogger twitter (why not anyone else?)
- f - Blogger Direct Recommendation
- g - Comment/Reader Recommendations
- h - Forum Discussion
Perhaps the most interesting thing about this list is that bloggers play a big part. I think it is an incomplete list with regard to every single viewer out there and every way of recommending.
Here’s probably how I would rank:
- a+f - I know a very small set of non-blogger anime watchers; I like bloggers.
- c - I like bloggers.
- g - If I get them, it happens.
- d+e - Quite passive on previews; just like seeing the pre-take. I casually skim tweets unless I’m there at the time, and it’s basically impossible to see impressions on one series among those I follow (without searching PLEASE!).
- h+b - Forums are messy, pro. anime reviews seem out-of-touch; I do like user reviews. (I have regard for some awesome old AnimeNFO user reviews)
So there are two sides of the situation, when we take a recommendation/suggestion and when we leave it, which is a bit interesting if we want to formulate an algorithm which maps this non-self influence of the experience. The benefit also has two sides answering the questions: 1) how does the viewer’s decisions on experience relate to recommendations? 2) what source has the greatest sway over listeners?
Developing a media-centric web-framework has it’s perks, and I may begin writing a solution in the near future; sooner if it seems immediately useful.
Notes
One of the features on melative is recommendations. Recommendations are distinct from Suggestions such that recommendations are for a given title, about the given title or from user to user, while suggestions are on a given title suggesting other titles of possible interest (notice the difference here).
Recommend: A user can recommend a title, in which a reason is given for experiencing said title, completely unrelated to another title. The title/reason can also be directed to another user, in which user1 recommends title to user2.
Suggest: A user can suggestion a title, in which a reason is given for experience title1, as a possible related interest to title2.
This is built into the old API and only partially transfered to the new API.
From this post, one interest might be ‘Reverse Recommendation’ in which a user can log that they accepted a recommendation in which there is no requirement for the recommendation to come from melative, but it should be categorized as a given source (from above) and about a specific media-title (ie. anime-Basquatch!). This could very well overlap the Associations engine mechanics.
I’ll leave it at that.
I’m remiss in building my melative account, but it’s something I’m very keen on doing sooner than later.
I see a lot of similarities in our ‘influencer profile’ and I’m very interested in how you go about answering the 2 questions you’ve presented.
I would not rank the sources because what matters the most to me is not who, but why.
@ghostlightning. dammit, I lost my reply O.o
Well, yea, building it up with all the experiences is a pain in the ass. Of course there could be some importing tools that parse lists from elsewhere, but I haven’t really found time for that. Also, having a complete experience isn’t really necessary for melative (it’s nice, but a subset works).
We’re exploring advantageous ways to interact with any form of media, so it could be used alone just for the relative ratings, the stream, recommendations, whatever. What matters is that interacting in almost any way on the site begins to build something… and if a user finds a really handy feature, there should be a way for accessing that data externally to the site, be it feeds or api or 3rd party apps/plugins.
@Kitsune, I like that. It sounds like, every recommendation treated especially. I’m not all that clear on the ‘why’ as a quantity or perhaps something like, the recommender knows one well and keyed in on some taste or idk?
[...] is explored further in ‘Influenced Path of Experience’ (Ryan A 2009/05/11). However, there’s something beyond this, that could explain the dramatic increase [...]