The Inescapable Algorithm? Part I
Several weeks ago Instagram announced that the photos of those users whom one follows will no longer just be displayed in a chronological order but will also be sorted according to their relevance. The change was however not to come into effect immediately but was to be gradually phased in. Nobody knows for sure whether their stream has or has not been changed already. I would call this a strategy of slow accustoming. A sudden launch may have led to more protests.
It is particularly interesting to learn what the impact of the new algorithm is and will be. Instagram claims that no user will miss out on the pictures of their friends. I have however made the following observation: In the late evening I usually post a picture. Oliver Warren, a film maker and photographer whom I got to know several months ago and whom I now follow on Instagram, also posts at the same time. Initially I did not notice that my stream no longer displayed new pictures by Oliver. This may be as I now follow over 600 people and I have stopped to realise immediately if somebody’s picture is missing. This would also only work if each user posted regularly and if it therefore caught my eye if there wasn’t a new one.
I logged onto Facebook and Oliver’s picture of the day, which he had posted some minutes earlier, was right there. I opened Instagram again to check that the picture really wasn’t displayed on Instagram. No, I couldn’t see it. Only when I called up Oliver’s profile I could then see his photo. After I had “liked” it, it began to be listed in my Instagram stream too.
Oliver and I have been friends on Facebook for a number of months. However, we have no other shared Facebook friends. Possibly that is why Instagram’s algorithm concludes that Oliver’s pictures are not relevant to me. But how does the Instagram algorithm determine relevance?
Fundamentally one can differentiate between formal characteristics and content-related features of pictures. Formal elements are the colour and the relationship between light and dark areas in an image. When I prefer dark pictures (and when I “like” them), will Instagram display such images for me? Is it enough to look at the photo or do I need to actively “like” it? Content-related elements relate to the question whether the photo is for instance meant to be taken seriously or in jest. A portrait can simply have different effects on different individual viewers. It is currently hardly imaginable that an algorithm can recognise and anticipate such effects. Or is it?
Another element determining relevance are likely to be the hashtags which can give hints about content. Hashtags are employed strategically and utilised as much as only possible – their information value is therefore small and, if in doubt, they say nothing at all. Under hashtag #Berlin you will not only find views of the city but also portraits of more or less good looking young people who like parading their bodies (or, on occasion, individual parts of their bodies) online.
Perhaps I should add at this point that I much like Oliver’s pictures and that his entire approach as a photographer fascinates me. He photographs people on the street from his hip level. I find it intriguing to guess in each picture whether it shows a person passing coincidentally in the street or whether it is somebody known to Oliver. Still, the algorithms are unlikely to be developed far enough to recognise why I find Oliver’s images exciting.