I want to understand behaviors of human beings through mathematical analysis of data. It may sound ridiculous to some people. Specially considering the “Interpretivist” movements in the discipline of behavioral science, perhaps, my research is not the right direction to move on. Despite that, I am interested on this topic due to the huge potential applications that I can see for mathematical models of human behavior. It is my way of staying foolish.
My background reflects more on why I am interested in this domain. I am originally an Electrical Engineer. Interest towards signals and systems is in my blood. However, I’m working in Human Computer Interaction where people are fascinated by both human and technology. So I think it is natural for me to attempt explaining humans using signals and systems. It is a difficult task, but becomes comparatively easier when we try to explain the data that humans leave behind. So I am interested to collect behavioral data about human and to explain it from novel or existing mathematical methods of signal processing.
This train of thought has a huge practical implication. Explaining human behavior would enable us to measure human qualities. For example, what might be a measure of good public speaking behavior, or a good husband’s behavior, or a good teacher’s behavior? May be one day we’ll be able to measure those just by analyzing data! If we can measure human behaviors we might be able to predict it too. Wouldn’t it be nice to automatically predict the impact of a presidential speech before actually delivering the speech? How about practicing and preparing for a job interview just with the help of a computer? We might be able to catch potential serial killers or greatest criminals from their behavioral data.
At this point, a question arise — Why machine learning (or computer vision)? Crowd-sourcing is currently being used to bypass AI problems. Why I’m not interested in solutions like that? A few reasons:
a) To me, crowd-sourced software looks like giving an answer without actually knowing the question. It is not praiseworthy because we are trying to solve artificial intelligence using artificial-artificial intelligence. Doesn’t it sound like cheating?
b) As I said, interest towards signals and systems is in my blood. My solutions might not be the perfect or the most user-friendly, but it will use the rigor of math — not the clairvoyance of human intelligence. That is a pleasure to contemplate for me.
c) Recent developments in machine learning, specially the developments in Deep Learning is enabling computers to do almost anything that humans can do. I do believe that deep learning can bring major break-through in medical science and clinical psychology. My research can be considered that kind of research in its infancy.
This doesn’t mean I don’t like crowd-sourcing. In fact, crowd-sourcing is one of the tools that I regularly use. However, I want to use the crowd intelligence to improve artificial intelligence, not to replace it. For example, crowd-sourcing could be used to collect massive ground truth data. But “using people to replace Google” — doesn’t look like a correct way to me. I like the concept of the people teaching machines for serving humans better, than the concept of converting people into humanized-machines.