Swinging Tags: A conversation with Koizuka Akihiko san

Koizuka-san is a programmer. No. This gives you the wrong idea. Better might be: Koizuka-san is the programmer. He created the original version of Nico Nico Douga in the course of 3 days, and has been the key person in its development ever since. To say it in his own words: “When I became a Nico Chuu, there was no other Nico Chuu”. The more technically minded Nico Chuu adore him. They credit him for the combination of complexity and simplicity that made Nico Nico Douga what it is today. A lot of the features of Nico Nico Douga show his hand writing. He programmed the comment function and he invented the unique version of tagging. Nico Chuus also value his approachability. I could experience this even before I had an interview by him: Koizuka-san became one of the first readers of this blog, and he bookmarked it from its start, so that other people would take notice as well.

"When I was a Nico Chuu, there was no other Nico Chuu" - Koizuka-san at Dwango office

In person, Koizuka-san is gentle and friendly, down to earth, precise, clear, logical and absolutely intelligent. When he talks you can feel clearly that you are listening to an outstanding brain. Even though he has a purely technical background, he often sounds more like an analytical philosopher. He was educated in the first cohort of the first high school in Japan that specialised in engineering. But even in this highly specialised technical environment he would often solve the tasks the teachers gave him, before they had stopped explaining the task. After his school he did not go to university, but became immediately a professional computer engineer, as it was believed at that time that a programmer has to quit his job with 35.

He is now beyond that age, but he still works for Dwango – a company that produces online entertainment content, mostly for mobiles, and is the parent company of Niwango (the company behind Nico Nico Douga). We meet him in the company and later we go for dinner. One of the things that impressed me most: I could feel the responsibility on Koizuka-san’s shoulders. Koizuka-san cares almost like a father for Nico Nico Douga and the community of Nico Chuus. Comparing him with a father might sound a bit odd, but it describes his position quite well: Of course he “only” created “generative algorithms”, and not the mad videos and comments on Nico Nico Douga, but he laid out the code, and since then he tries to develop and protect it.

Part of our interview is about the early days of Nico Nico Douga. Some of his stories are thrilling as suspense movies. One sticks to my mind particularly: The dramatic moment, when first the comment server of Nico Nico Douga broke down, and then, right at that moment, Youtube cut them off without notice. In the following days, the programmers of Niwango installed a new infrastructure, which allows Niwango to stream all videos by themselves. To do such a thing in days is quite an incredible technical achievement. By the way, up till today, Nico Nico Douga combines three different forms of servers – the web server provides the interface, a video servers streams the video data, and a third server provides the layer of comments.

Koizuka-san gives a paper on “How to create Nico Nico Douga” on a conference (“Itpro Challenge!”), then uploaded on Nico Nico Douga. Koizuka-san finds it generally strange to read comments about himself on the videos, but, as a 2channeller, he says he can take it.

Another early-days-story is the creation of the comment function on Nico Nico Douga. In this discussion we start to analyse the different levels of time that are incorporated in the original interface: The time of the video, the real time of the viewing, the time of the writing of the comments, the time of the scrolling and reading of the comments, the time of the archive of the comments (which is still not a real archive, as only the last 1000 comments stream through the video). Nowadays, these are not the only forms of time: You have also the time of the tags, the time of the evolving genres, the time of the user feedback, the time of the ranking … and Nicoscript has added even further layers of time, for example additional layers of automated feedback, while the users are watching the video. That the whole things still remains its natural and intuitive feel is almost a miracle and largely an achievement of the work of Koizuka-san.

I originally thought that the comments are limited because otherwise the screen would become too littered. This is the case, but there is more to it: For Koizuka-san the main reason was to keep the experience simple. If there are too many comments directly accessible for the user, the users might get scared, as they would theoretically have to know all the old comments, before they can add their own comments. Such old comments might refer to old forms of commenting or old references, that new users might not yet know anymore. The limitation of comments keeps them fresh and simple, and users can comment directly and affective. This is just one example of Koizuka-san’s style of programming: Simple ideas that make things simple for the user. Yet, as we all know, there is nothing as hard as this simplicity.

That comments scroll from right to left though the video, was decided from the start. However, how they would scroll, at what speed, and on which place in the video, was developed by Koizuka-san and his team. Scrolling is one way how the experience of Nico Nico Douga is kept simple and user friendly. As there are only a small numbers of comments displayed in any given moment of time, users can add their comments almost as if they would add a comment to a thread with one simple layer – the layer of the general tendency of the comments at any given moment of time. This form of a simple pseudo-thread-structure enables the users to read the comments and to react to them intuitively and directly.

A further simple, yet highly effective idea is the form of the daily ranking. It is based on how many users have added the video on their “Mylist” on the day. Mylist is Nico Nico Douga’s form of bookmarks. Users put videos on Mylist usually after they have seen it and when they plan to see it again. You can add each video to your Mylist only once. There are other possible forms of ranking too (according to the number of comments, for example), but Nico Chuus use the daily ranking based on Mylist as the first thing to go, when they check Nico Nico Douga. The ranking is so accepted, that Nico Nico Douga has re-gained a function that used to be exclusive to TV: When Nico Chus meet, they can discuss the videos on the ranking of that day, just as TV watchers can discuss what has been on TV yesterday evening.

I was curious to find out how videos make their way on this list directly after they are uploaded. Freshly uploaded videos appear for a short moment as “new arrival”, but afterwards they are in danger of drowning in the database. To prevent this, many uploaders ask their friends to add their freshly uploaded video to their Mylist, often via Mixi (an interesting dimension of social networks in the otherwise anonymous world of Nico Nico Douga). Until the video becomes known, uploaders have enough time to add their own comments. This is important because the first comments influence the later kuuki of the video. The freshly uploaded video is precious, almost like freshly fallen snow. Some users hunt these unknown videos. They want to comment first, and afterwards they have an interest in promoting the video. If all this does not work, the video can still generate a specialised audience through tags – and this specialised audience can then be the basis for a general success in the daily ranking.

Maybe Koizuka-san's most important achievement is the idea to limit tags to the amount of 10. I have already been raving quite a lot on this blog about the geniality of this idea, so I keep it short at this place. His original aim was to reduce the load on the database and keep the system simple to use, but he soon realised that there was much more to it: Limited tags, as I have written before, are key to the collective negotiation of the evolution of genres and subgenres. For Koizuka-san, tags are not only about description, but also “viewpoints”. In the beginning, Koizuka-san was concerned, when he noticed that tags“swing”on Nico Nico Douga: Competing viewpoints are set against each other. Koizuka-san spent quite a lot of time to develop a program that would “solve” this problem, until he realised that the collective intelligence of the users does indeed do the job better then any program can: At some point, one of the tags wins, and “sucks in” and therefore incorporates the tag that stood for the other side of the swinging movement. Hegel would be excited!

At the end of our long and at every moment fascinating conversation we started to think about how Koizuka-san might ‘copyleft’ his ideas. Although Koizuka-san is by now pretty famous and has entered the ranks of the programmer-stars of Japan, neither he as a person nor his ideas are known outside of Japan. He is keen to share his experiences and inventions, and I think it might become one of the tasks of the metadata project to communicate them. Some of these inventions play on a Japanese background, but they might still be a starting point to re-think our Western ideas around folksonomies or around the semantic web. This will need more space then what I can describe in the limits of one blog entry. But I hope this corporation can go on.