Computational-level analysis of insight problem solving
No Thumbnail Available
Having an insight is a central aspect of the human ability to do problem solving. When trying to reach a solution for a problem, having an insight is what lets us formulate a problem in a way that we can come up with an answer. However there is no scientific consensus about the cognitive mechanisms of reaching insight. In this thesis we will briefly present the current literature about insight problem solving and show some of the shortcomings that have been affecting it. In particular, we will ask if existing models of insight problem solving are able to generalize and be applicable to all insight problems or only to a selected few they are specifically designed for. We will focus on an existing model of insight problem solving, and give a computationally-backed argument about how the existing model could theoretically be used to encode any type of problem. We will also give an example of an encoding of an insight problem in the existing model and from there we will point out that the model seems to construct its input so that the solution is easily found, almost built in to the model. We will argue about the risks of this in-building and then we will consider what could be a minimum to be built-in. This will lead to a reformulation of a more useful model capable of insight. Finally, with another result we will show that an important aspect of problem solving is often overlooked, namely the amount of time (or steps) necessary before finding the solution. Indeed we will show that pre-specifying this amount is actually necessary for a cognitive model of insight. These results will give important theoretical constraints to future theories of insight prob- lem solving. Furthermore, the thesis will suggest specific future research approaches for ad- vancing our knowledge in this field.
Faculteit der Sociale Wetenschappen