“Algorithm”, “Recommender systems”… You know these words but well, I’m pretty sure this is Greek to you. Indeed, the technical complexity make them hard to understand and also to identify.
What is a recommender system? It is quite easy because we all know some: Amazon’s “you should buy”, Netflix’s “you should watch” / “here’s the next movie”, Spotify’s “you should listen” / “here’s the next song”. You have the big picture.
A recommender systems is an algorithm that receive data in input and give a result in output. Let’s talk about algorithms:
Data = ingredients and output = cake. From data to output, you have to mix/spread/bake/test/remove from oven/let cool. Easy!
Okay, it’s quite simple here to understand the “rece
ipe” because you have 6 ingredients, 6“tasks”. You also know what to get on the other side.
But here is the problem: You’ve baked this cake and you bring it to your friend’s birthday. He loves the cake (houra) and would like to do the same one. The solution is: give him the recipe and that’s it. But no, you don’t. The cake’s recipe is a family-secret!
Your friend is confused but he doesn’t give up. He knows he will never bake the same cake but he can do a quite similar one.
So, he starts to ask you about the ingredients. Alright! You tell him about the chocolate, the sugar, the flour (obvious), the butter and the eggs. But nothing about the vanilla :)
Now he’s asking about the tasks to follow. You answer “bake, let cool, mix, spread”. Well, he thanks you… but these news are not a scoop. He wants to know how much of the ingredients you need and the exact tasks to procede. Impossible, this is THE secret. Your friend is in front of the cake black box.
Anyway, everybody at the party eat the cake and has fun. You’re the last to leave and before you go, your friend is asking you some question, not about the “how to”, but the “how you do”. Questions you can easily answer because they not reveal the recipe’s secrets.
You start to talk about the history of the recipe, how you did it for the first time, how you try to get a better cake, what do you use to bake it, which kind of butter you buy. In fact, he’s doing a “cake ethnography”. He don’t want to (or can’t) have the core informations but he thinks it’s even more interesting to get the “insights” of “cake baking”.
In my internship, here in Berlin at the Centre Marc Bloch and with the Algodiv team, I’ll become your friend and lead an ethnographic study (not on cakes baking, no worries, but on recommender systems for web-services).
What is the Algodiv project ?
“Algodiv is an interdisciplinary research project focusing on the question of information diversity in online communities and the web, with an additional focus on the effect of algorithms and algorithmic practices.
Funded by ANR from 2016 until 2019, the consortium essentially gathers sociologists and computer scientists who will adopt a mix of qualitative and quantitative approaches on a variety of web environments, from blog and micro-blog ecosystems to social media platforms, through wikis and image- and article-sharing websites.” (source)
I would like an will try to do my best to write at least one article a month here on Medium to describe and discuss my work and process. Feel free to comment :)