Social Sense Computing

I chatted with wan-rong about intelligent home appliances this afternoon. We talked about such questions, "How do people know that someone is going out?" "How do people know that someone is busy?" The answers to the first question may be, "dressing" "carrying a bag" "carrying the key" "taking a shower" "waring socks" and so on, but none of them are necessary conditions. It is absolutely different when someone is going out for a dinner and going to 7-11 to buy some snacks, and all these ques can either take place or not for the different conditions. The answers to the second question may be "He/She keeps working even though seeming to be tired", "He/She do a lot of things in a hurry," and so on, but it depends heavily on the situation as well.

How is the enviornment gonna be smart? So many researchers have done lots of work on this question, while none of their works have persuaded me that it would be really nice to have all these home appliances in my house. The truth is that, I'd rather use tools with no intelligence at all, and I believe many people have the same feeling too. Why? because there're so many things that "smart" home appliances don't know today. The troubles brought by these smart systems are thus so more than the problems solved. Researchers in various areas have tried to make the systems even closer to how we people think and even feel about things, such as commonsense reasoning and affective computing, but I think there's something important that is still not addressed -- social issues. In order to make these systems truly close to humanity, it is important to the home appliances to apply the knowledge about how people think of, and observe others' behaviors. Such knowledge hasn't been incorporated in commonsense computing and affective computing literature. Affective computing techniques, as in many other AI technologies, are closer to psychological ways of dealing with humans' behaviors, instead of sociological ways. Commonsense, on the other hand, is more on the phenomona of how things generally appear, and is superficial and less specific about humans' interactions. The important rules of how people get together with each other have been missing in to-date AI and context-aware technologies, making the "smart" environment still dumb and hard to get along with.

I would say social sense is like one of the sides of a triangle, and commonsense computing and affective computing are the other two. The three components cooperatively make up the reasoning mechanism that computing devices or environments can apply to understand, and maybe to benefit, people's lives. We agree with the idea that humans' social behaviors have not been applied widely in AI or other fields because it is originally hard to conduct sociological experiments. Unlike most of the psychological experiements, it has to be done with more than one subjects. It thus becomes difficult for computer scientists to come up with an good data collection mechanism for dealing with such knowledge, not to mention knowledge representation and computational models. "But at the same time, this is the very reason why it is so interesting." said wan-rong. I agree with her very much.

Nevertheless, I don't think it is not impracticable at all. Maybe the way of constructing commonsense computing infrastructure is a good example of how we can do for people's social sense - making a website, having people key-in their ideas of how people interact with each other in different social settings, and making up a graph representing all the collected knowledge. It would probably be different from commonsense in many aspects, since the data is more dependent on the situations, but I think it would be pretty similar. It is also more difficult to illlustrate that this idea is correct through applications, since it's more into the real life, and more into the physical space that we live in. But in order to reach the holy grail of making people's lives truly easier with the aim of technologies, I think we should give it a try.