User Interfaces for Human-Algorithm Collaboration through Modelling

点击翻译按钮获取中文摘要

My research program is in the area of Human-Computer Interaction and intends to allow people who do not have mathematical or computer science backgrounds to take advantage of the large existing knowledge of algorithms when addressing their own personal problems and questions on their data. We can find computational solutions to many of the problems that people come across in their professional and personal lives. For example, consider a family who is trying to coordinate transportation for their kids' extracurricular activities (how can we minimize costs while reducing waiting times?), or a Grade 8 teacher who needs a seating arrangement for a class (how do we sit pupils with complementary strengths close to each other?). Unfortunately, it takes expertise and time to be able to formulate the specific problems in a way that the computer can solve them. People are unlikely to hire an expert or learn how to program a constraint programming language in order to find a better solution for something that is immediate and very specific to their current circumstances. Hence the problems stay unsolved, or we have to cope with mediocre solutions and, for some of us, the nagging feeling that this can be done better. This research is a quest in search of better ways for the algorithms and the people who need them to communicate with each other. In other words, I want to improve Human-Algorithm Communication (HAC). If people could, perhaps with help from the computer, formulate their problems in quick and efficient ways that are sensitive to the context where the solution can be worked out and applied in, we would not need experts or lengthy training to get better solutions. Simultaneously, if the outputs and parameters of the algorithms can be communicated to people in improved ways, people will be more likely to trust calculated solutions and to make use on the vast existing knowledge that we have on algorithms. The research program involves the study of human problems, the context in which they arise, how they are identified, and how humans conceptualize them. It also involves the design of new description languages to express those problems, and building interfaces that support good human-algorithm communication. Success will mean that more problems will get better solutions, for a potentially very large number of people who, currently, do not have ways to take advantage of this hard-won (and largely tax-payer-supported) existing knowledge. Because we will start work targeting a specific section of the professional population (administrators and administrative assistants), people in these jobs should be soon able to benefit of improved productivity and more challenging and stimulating work environments. The knowledge gained will also help build better user interfaces for sophisticated tasks, which will benefit other researchers and companies who build software tools for this populations, in the local economies and in Canada at large.