Final Theses

People are often confronted with decisions between options that differ on multiple dimensions: one dish at the mensa is super delicious, but not very healthy, another one is probably the best for my body, but I had it already yesterday,… 
Also in decisions between options with uncertain outcomes, for instance when planning a trip with the Deutsche Bahn, some outcomes may be desirable (everything going smoothly and you arrive on time), and undesirable (waiting at a station for the next train for two hours). 
We study this kind of decisions in highly controlled environments and with highly controlled stimuli in the lab (e.g. using risky choice; decisions between lotteries with described outcomes and probabilities). Attention plays a crucial role in decisions in that it seems to push us towards an option at which we look for a longer time. In addition, attributes (like taske or healthiness) get more important, the longer we look at them. Mathematical models are used to explain this relationship. 

This project should answer an important questions: Is the effect of attention different when we look at positive attribute values (e.g. potential gains in lotteries; or the tastiness of a dish), compared to subjectively negative attribute values (e.g. the *unhealthiness* or a potential loss). 

This project is well-suited for groups of two to three students because of the various different aspects. Students should have good statistical background and be familiar with the programming language R. 

Supervisor: Sebastian Hellmann (sebastian.hellmann@tum.de)

Some initial readings:

• https://elifesciences.org/articles/103736
• https://link.springer.com/article/10.1007/s10683-024-09849-7
• https://www.sciencedirect.com/science/article/pii/S0749597821000431

This project aims to investigate the prevalence of magnitude sensitive behaviour in risky decision making and explores potential accounts for this behavious in common theories of decision-making. 

Magnitude sensitivity describes the effect that, when deciding between two options, making both options better leads to faster decisions. Whether decisions are more or less accurate seems to depend on the specific task. Magnitude sensitvity is found across a wide range of decision tasks from perceptual decisions (e.g. deciding which of two gray patches is brighter; which of two food snacks is more preferrable,…) and there are different possible explanations for this mechanisms. 

The effect has not yet been studied in the context of risky choice, i.e. chosing between two lotteries with different outcomes and probabilities. These decisions are commonly investigated using computational models like cumulative prospect theory. In the special case of risky choice magnitudes can be increased in different ways: First, by increasing the potential gains (positive outcomes of the lotteries), or by increasing the probability of high-reward outcomes and lowering the probabilities of low-reward outcomes. These manipulations have different effects on the decision to be made, according to CPT. The project aims to systematically describe the effects that magnitude manipulations have on the internal decision process and to investigate whether according behavioral patterns can be found in human data. 

Supervisor: Sebastian Hellmann (sebastian.hellmann@tum.de)

Some initial readings:

• https://link.springer.com/article/10.1007/BF00122574
• https://www.nature.com/articles/s41562-020-0822-0
• https://link.springer.com/article/10.3758/s13423-017-1383-8
• https://link.springer.com/content/pdf/10.3758/s13423-015-0858-8.pdf

Loss aversion is one of the most prominent concepts in the study of risky decision making, describing the widely observed phenomenon that people tend to assign more subjective value to losses than to equally sized gains. Recent research has demonstrated that loss aversion may however have boundary conditons. For example, the absolute magnitude of outcome values seems to affect the degree of loss aversion, leading to debates if established formalizations of prospect theory models should be adapted. The focus of this thesis project is to implement a series of functional forms of loss aversion's magnitude dependence to systematically compare their model fit in a Bayesian hierarchical modeling approach.

Requirements: Experience in data analysis using R. Of advantage: Previous experience with computational modeling

Supervisor: Nuno Busch (nuno.busch@tum.de)

Recommended Literature:

• Mukherjee, S., Khan, O., & Srinivasan, N. (2025). The role of magnitude in loss aversion. Decision. https://doi.org/10.1037/dec0000256
• Harinck, F., Van Dijk, E., Van Beest, I., & Mersmann, P. (2007). When Gains Loom Larger Than Losses: Reversed Loss Aversion for Small Amounts of Money. Psychological Science, 18(12), 1099–1105. https://doi.org/10.1111/j.1467-9280.2007.02031.x
• Ariely, D., Huber, J., & Wertenbroch, K. (2005). When Do Losses Loom Larger than Gains? Journal of Marketing Research, 42(2), 134–138. https://doi.org/10.1509/jmkr.42.2.134.62283
• Mukherjee, S., Sahay, A., Pammi, V. S. C., & Srinivasan, N. (2017). Is loss-aversion magnitude-dependent? Measuring prospective affective judgments regarding gains and losses. Judgment and Decision Making, 12(1), 81–89. https://doi.org/10.1017/S1930297500005258
• Novemsky, N., & Kahneman, D. (2005). The Boundaries of Loss Aversion. Journal of Marketing Research, 42(2), 119–128. https://doi.org/10.1509/jmkr.42.2.119.62292
• Zeif, D., & Yechiam, E. (2022). Loss aversion (simply) does not materialize for smaller losses. Judgment and Decision Making, 17(5), 1015–1042. https://doi.org/10.1017/S193029750000930X
• Hof, L. & Busch, N. (2025). Bayesian Cognitive Modeling Tutorial. Github. https://github.com/linushof/TutorialsForTeaching/blob/main/BayesianCognitiveModeling/bayesian_cognitive_modeling.md 

In experience-based decisions, such as in the sampling paradigm, people have no initial knowledge about possible outcomes and their probabilities of each option. Instead, they can explore their option first in a sampling phase by drawing exemplary outcomes from each option, before committing to a final consequential choice. Reinforcement learning frameworks model this process as sequential updating of expectations for each option during information search. Importantly, it has been shown that people update their expectations differently depending on whether the sampled outcome is a positive or negative surprise (prediction error) relative to their current expectation. This is called asymmetric learning. In recent research, we have found asymmetric learning from gain and losses in monetary lotteries. The present project seeks to dissociate the asymmetries in learning between outcome domains (gains and losses) and prediction errors (positive and negative surprises).

Requirements: Experience in data analysis using R. Of advantage: Previous experience with computational modeling

Supervisor: Nuno Busch (nuno.busch@tum.de)

Recommended Literature:

• Gershman, S. J. (2015). Do learning rates adapt to the distribution of rewards? Psychonomic Bulletin & Review, 22(5), 1320–1327. https://doi.org/10.3758/s13423-014-0790-3
• Hertwig, R., & Erev, I. (2009). The description–experience gap in risky choice. Trends in Cognitive Sciences, 13(12), 517–523. https://doi.org/10.1016/j.tics.2009.09.004
• Hof, L. & Busch, N. (2025). Bayesian Cognitive Modeling Tutorial. Github. https://github.com/linushof/TutorialsForTeaching/blob/main/BayesianCognitiveModeling/bayesian_cognitive_modeling.md
• Palminteri, S., & Lebreton, M. (2022). The computational roots of positivity and confirmation biases in reinforcement learning. Trends in Cognitive Sciences, 26(7), 607–621. https://doi.org/10.1016/j.tics.2022.04.005

Study Programs:

MMT: Master in Management and Technology

MCS: Master in Consumer Science

BMT: Bachelor in Management and Technology

BBWL: Bachelor in Technologie- und Managementorientierte Betriebswirtschaftslehre

Upon agreeing on the research question that the theses should answer as well as a detailed research plan, supervisors register the final thesis via Koinon. Students will be automatically notified via e-mail by the system and are asked to confirm the registration. After the registration, students must submit their work within 3 or 6 months.

Students submit their final theses (PDF document) as uploads via Koinon. After students submitted their thesis in the portal, supervisors will be automatically notified via e-mail by the system and can access the thesis via the portal. 

Master's theses should be written in English. Bachelor students are also encouraged to write their thesis in English; however, after consulting the supervisor, a Bachelor thesis can also be written in German.