Topics
Target group: | M.Sc. (Computer Science) |
Description: | Bayes-in-the-head theories of cognition assume that humans calculate or approximate Bayesian statistics for everyday reasoning. Other approaches, however, assume mental representations with a qualitative ordering on models. From a formal perspective, it is an open question which of these approaches predicts human reasoning behavior the best. The goal is to compare, implement and evaluate the theories against empirical data. |
Requirements: |
|
Target group: | B.Sc./M.Sc. (Computer Science / Cognitive Science) (several theses) |
Description: | For given information what inferences do humans draw in general? Going one step further: Is it possible to predict for some background knowledge like working memory size and inferences drawn for some problem instances what the subject draws for an inference for a similar problem? What about a not so similar problem? |
Requirements: |
|
Target group: | B.Sc./M.Sc. (Computer Science / Cognitive Science) |
Description: | Although forgetting is often associated with frustration in our everyday lives, forgetting is a highly adaptive mechanisms that helps us to deal with the great amount of information we are confronted with on a daily basis. In fact, people are even able to intentionally forget previously learned information. Yet, the cognitive underpinnings of intentional forgetting are not yet well understood. We are currently running several experiments to gain additional insights on how people are able to intentionally forget different kinds of information (e.g., semantic memory or motor representations). Furthermore, we are developing a cognitive model that can account for such mechanisms. At the same time, we aim to use data-driven approaches to obtain upper bounds of predictive performance for these models. |
Requirements: |
|
Target group: | B.Sc./M.Sc. (Computer Science / Cognitive Science) |
Description: | Decisions human make can depend on heuristics. In this thesis you will start with analyzing existing heuristics and then try to automatically generate heuristics that can fit better human decisions than existing ones. |
Requirements: |
|
Target group: | M.Sc./B.Sc. (Computer Science / Cognitive Science) |
Description: | How can we compare different cognitive theories? Mathematical psychology offers interesting approaches to evaluate theories using AIC, BIC, DIC etc. Multinomial process trees offer an excellent possibility to formalize cognitive theories (especially for research on recognition memory. Their expressiveness, however, and their application on human reasoning still requires more research. |
Requirements: |
|
Target group: | B.Sc./M.Sc. (Computer Science/Cognitive Science) |
Description: | This project is about the modelling of reasoning processes in the Neural Engineering Framework (NEF). The NEF is a cognitive architecture framework which is based on physiologically plausible neural clusters. The goal is to implement a model which is able to solve reasoning and tasks and afterwards to evaluate the results and the model. In the end, we would like to find out how algorithms have to be implemented in neurons in order to exhibit human-like behavior. |
Requirements: |
|
Target group: | M.Sc. (Computer Science / Cognitive Science) |
Description: | Problems such as the Winnograd challenge or PDP-Problems are considered very hard AI problems. Such problems are typically very easy to solve for human reasoners but are very difficult for current Machine Learning approaches as they require some form of semantic understanding. |
Requirements: |
|
Target group: | B.Sc./M.Sc. (Computer Science / Cognitive Science) |
Description: | The syllogistic reasoning domain stretches over 64 distinct problems. As a result, experiments testing participants on all instances take a significant amount of time resulting in a potential loss of data quality due to fatigue, boredom, or a general loss of concentration. This project investigates the impact/influence of specific syllogisms on predictive cognitive modeling with the goal to propose a reduced set of problems balancing data quality and richness. |
Requirements: |
|