Humanising AI

At KTH, if you keep your eyes wide open, your schedule will never be free. Wait that does not sound good, but it is! I am not talking about projects, tests and exams (although they exist, of course), instead of about events.

Yesterday there was a new one: Humanising AI, a two-day event about how AI and HCI interplay. It was great!

What Was The Event About

Think about everything that is run by algorithms: your email box, your car or bike route, or news feed, maybe. The fact is that Artificial Intelligence is changing the world, and of course the fields of computer science and human-computer interaction. But how can designers build better systems for this new environment? This is the core question that this event wants people to think about, not solve, as the task is humongous, but to reflect.

The main organizers were Kia Höök, a great professor from KTH, and Barry Brown, from Stockholms Universitet. They two were also involved in the Mobile Life Center, what you might recall if you have been following this blog for a while. Anyway, they decided to organize the event having lots of talks on the first day, and a workshop on the second one.

The Talks

Unfortunately, I could not get tickets for the workshop, so I went only to the first day. Here is a little about how it went.

The first talk was given by Albrecht Schmidt, a professor from Ludwig-Maximilians University, Munich. He talked about how AI can amplify the human mind and we can benefit from more advanced computing.

An interesting fact he brought presented is that the average US house has around 300 thousand objects. Now, what happens when every one of them is a small computer, connected to one another? This is where things get interesting. Professor Schmidt also mentioned that everyone is looking into how to automate big things (cars, cities), but that the big interaction questions are in the small ones, on everything that we use without even noticing (doors, clocks, keys).

Finally, an interesting inquiry was about information overload. According to him, it is a design question, not a psychological one. Users are overloaded because things are badly designed, not because that is an excess of information.

Then, Professor Susanne Bødker, from Aarhus University, took the stage. The main concept she introduced was the one of “Run Away Objects”, that are objects that are changed and modified by many users but not owned by any. Think about open source software, or maybe cooperative platforms like Wikipedia. Given lots of current platforms are automated and based on automation, how can we maintain control of them? That was her question.

And the third presentation of the first panel was by Karl Johnsson, a researcher from KTH itself. His field of research is transportation automation, something he has been working for the past decade. He presented a project where the researchers partnered with Scania to design a plateau system for trucks, that is, for vehicles to automatically enter in line to save fuel. The dream goal would be to implement that in Continental scale, which would mean saving tonnes of fuel and money.

It was then time for a small coffee break. Relax and chat before the second panel session, opened by Barry Brown, who I mentioned before.

Barry talked about the relationship between HCI and AI. He mentioned an old paper, where the author argues that both fields have the same goal (to promote a better user experience) but differ in approach. While HCI focus on the human side, AI focus on automation.

After Barry, came Antti Oulasvirta, from Aalto University, in Finland. Antti talked about user interfaces in an age of AI, which was really, really interesting. He talked about computational thinking as a way to approach design problems, which would include breaking down the problem into minimal tasks that can be automated. This can lead to a mathematical approach to design optimization, and Antti also showed some projects on which software would evaluate different design decisions to provide the ideal one. For example, a software to create and evaluate different positions of images on a website and give the optimal for a specific goal.

If Antti was highly computer-minded, the talk after him was completely different. It was professor Benjamin Shestakofsky, a sociologist from Berkeley California. He talked about his Doctoral thesis, an ethnographic study he did inside a startup. Hist talk was interesting due to the fact he analyzed automation and AI, and showed how organizational and cultural factors are essential for the adoption or not of new technology.

And finally, the last talk I managed to listen to (then I had to leave to run some errands) was David Shamma, a data scientist from Palo Alto. He has a background working at Yahoo and Flickr, developing machine learning algorithms. He talked mostly about his experience and past projects, and why lots of AI ideas go wrong. It was interesting because it felt more like an “industry insider” perspective.

Take Away

Once again I am surprised by these events at KTH. You see if back in Brazil we had an event with people from such Universities and Research Centres like this, it would be a huge thing. Here it feels way more casual, where you can actually get in touch with professors and teachers. It was great!

Hope you liked it,

/tomas

 

Leave a Reply

Your email address will not be published.