Final Theses
The field of empirical social research has a wide variety of social problems and issues that can be exciting as topics for a sociology thesis.
In general, we are most interested in supervising research in the areas of social network analysis, agent-based modeling, computational social science, and applied quantitative empirical social research. You will find a list of suggested topics below.
Social Networks
Friendship paradox: Are your friends more likely to be healthier and happier than you?
Induced centrality and “importance network”: Who is important for whom?
No kidding: Are jokes all relative?
Thresholds, latency and complex contagion: What happens when some people just need more time to make up their minds?
Wisdom of the crowd: How do different social influence mechanisms affect group diversity and the wisdom of the crowd?
Overconfidence and diverse problem solving in group Social networks and learning: How does knowing your team-mates improve interaction?
Mental health and perception of social relationships
Codevelopment of personality traits and social networks in adolescence Negative ties and loneliness
Agent-Based Modeling
Multidimensional segregation
Luck vs talent: Why are the luckiest people more likely to be moderately talented?
Other
Social status and life expectancy: Do you live longer when you win a medal at the Olympics?
Broken window theory: Do individual level attributes moderate the spreading of disorder?
Say “ cheeeeeeze”! The emergence of smiling faces in advertisements
Detailed description of the topics
Social Networks
Friendship paradox: Are your friends more likely to be healthier and happier than you?
In some previous research, I showed how your friends not only tend to have more friends than you have but are also more likely to have certain attributes (see Grund 2014 ). The random friend of a random person is not random anymore. This project investigates social network data from school classes around Europe and applies this logic to different kinds of health-related attributes. Is it indeed the case that on average your friends are healthier and happier than you?
Induced centrality and “importance network”: Who is important for whom?
Everett and Borgatti (2010) wrote an interesting paper on induced centrality. This concept captures the notion that individuals in networks are not just important by being connected well, but rather by making others important. Induced centrality is then calculated by temporarily removing an individual from a network and investigating how this removal affects the importance of others. More generally, this perspective on centrality/importance in networks taps into sth very interesting. Not everybody is equally important for everybody else. For example, the most important person in your life might be somebody who is not that well connected, but that person is important for you. While Everett and Borgatti (2010) looked at this at an aggregated level, one can think of an “importance network” instead, which captures for each pair of actors A, B how important these two actors are for each other (within the framework of induced centrality). This project develops the R routines to calculate such “importance networks”, elaborates on the concept and then applies it empirically. Maybe we can see that relationships that are important in the “importance network” are also stronger or substantively different?
No kidding: Are jokes all relative?
This project builds on and empirically investigates the anecdotal observation that we make jokes in a relative way. Would people tell the same joke differently when they tell it to a different person? And which characteristics of both joke teller and joke listener matter here? To study this we use a vignette design where different scenarios about teller and listener are given and then participants are asked to fill in blanks. For example, “Three persons meet. The first one is from ___, the second one from ___ and the third one from ___.” What would participants of a study fill in here? Based on this information we will then also map out a network of joking. Who makes jokes about who? Are there any countries or regions who get joked about a lot and how is most central in this joking network? Seriously, all serious research - no kidding.
Thresholds, latency and complex contagion: What happens when some people just need more time to make up their minds?
The seminal paper by Granovetter (1978) introduces the ideas of thresholds, where individuals start acting when a certain amount of other people already act and their threshold is met. However, this paper implicitly assumes that action follows instantaneously after a threshold is satisfied. There is no notion of time delay, latency or deliberation in there. In reality, social action often requires individuals to think about stuff for a bit and process things. While the psychological literature acknowledges differences of individuals in such response times, the sociological literature on social diffusion does not. This gets very interesting when we are talking about complex contagions (see work by Da,pm Centola), which essentially says that social contagion is different from e.g. viral contagion in the way that people need more than one infection at the same time (I only start doing sth when I see more than one of my friends doing something). Although social contagion does make a lot of sense, it could be overrated simply because people have different response times, i.e. it looks like an individual (focal actor) only started acting after two of her friends acted, but in reality one friend already sufficed to make her start but the focal actor just took a bit longer to respond and in the meantime another friend of hers acted, which gives the impression that the focal actor only started because at least two friends acted. A project could elaborate on this idea more and investigate under which conditions how much contagion that we assume to be social is in fact not. The project would simulate the diffusion (in different network contexts) and combine it with different response times of individuals. Next, it would analyze the simulated data as if there would have been social contagion (while in fact, we know there has not because that is how the data was simulated) to quantify the amount of social contagion we would detect while in fact it is not there.
Wisdom of the crowd: How do different social influence mechanisms affect group diversity and the wisdom of the crowd?
Hong und Page (2004) present a model that shows how group diversity (and the ability to draw on more different heuristics to solve a problem) can collectively lead to higher outcomes. A simple idea for a thesis project would be to re-implement their model and extend it by adding social influence effects. What happens when individuals copy the strategies of the individually more successful individuals? What happens when they copy the strategy of the least successful individual and so on? In fact, one can show that while it makes for each individual sense to copy the heuristics of the most successful individual to optimize their own individual benefit, collectively this is not the best strategy .
Overconfidence and diverse problem solving in group
Hong and Page (2004) present a model that shows how group diversity (and the ability to draw on more different heuristics to solve a problem) can collectively lead to higher outcomes. Their model, however, implicitly assumes that agents contribute equally to solving problems. In reality, however, some agents are more confident than others to solve problems based on their previous performance. Such confidence most likely matters when group members decide on which heuristic to apply next to solve a problem collectively. At the same time, an overconfidence of some agents leads to a reduced set of heuristics being applied. Hence, although groups might in theory have a diverse set of heuristics to choose from, the overconfidence (and consequential dominance) of some group members effectively reduces the diversity of heuristics the group actually uses. A simple idea for a thesis project implements a simple version of Hong and Page’s model and implements the idea that groups do not randomly select heuristics among their midst, but rather give preference to individually successful heuristics. This could lead to the finding that although diversity in groups is important, it is actually the group’s capacity (or indifference) to experiment and cycle through the diverse set of heuristics available to the group. Obviously, this creates some interesting tension with Becker et al.’s paper on the wisdom of the crowd where confidence actually improves group performance. Based on Hong and Page, I would hypothesize, confidence is actually detrimental to group performance because it reduces diversity of heuristics.
Social networks and learning: How does knowing your team-mates improve interaction?
This project investigates the importance of team learning (or what I previously called relational experience, see Grund 2016 ) for the successful interaction of soccer players. It builds on this work of mine: https://www.youtube.com/watch?v=3l_HY8GwBm4 . The project will be literature only to lay the foundations for an empirical analysis I previously conducted, which showed that when football players know each other for longer, they are more likely to successfully pass the ball to each other. In fact, I also showed that this effect is moderated by the individual quality of soccer players.
Mental health and perception of social relationships
How reliable is the recall of social relationships? A symptom of many mental illnesses is the distorted perception of social relationships, which can jeopardize the maintenance of those very relationships. While the influence is well studied for the immediate social environment, it remains unclear to what extent the social structure of the extended social environment may also be perceived distortively. This thesis aims to investigate how mental health affects differences in perceptions of social relationships among other individuals. The thesis may take the form of a systematic review or a research design.
Codevelopment of personality traits and social networks in adolescence
In personality research, the importance of genetic predispositions and related genetically driven maturation is often emphasized and personality traits are considered to be mostly stable over time. On the contrary though, longitudinal empirical research suggests that changes in personality appear over the whole life course. Especially for adolescence and early adulthood empirical findings indicate the occurrence of maturing processes and that some traits tend to decrease and others tend to increase. These findings strengthen the importance to investigate which factors trigger changes in personality traits. In adolescence, the social influence of peers is of particular importance. Furthermore, it can be expected that positive or negative sanctions as responses to the display of one’s current personality traits are of importance for further development. Personality traits also influences the individuals' social relationships and how these are perceived. Depending on their personalities individuals can have different responses to the same stimuli. Individuals with e.g. high levels of extraversion tend to perceive the same social interaction as more pleasant, resulting from their larger extent of sociability. Shy and reserved behavior may lead to a smaller number of social contacts. The thesis should explore the codependence between personality trait(s) and social relationships in adolescence. It may take the form of empirical work or a systematic review.
Negative ties and loneliness
Loneliness poses an acute and widespread threat to physical and mental health in the Western world. Social isolation and loneliness increase the risk for early mortality, coronary heart disease, and dementia as well as mental illnesses. Social isolation refers to the actual and objective characteristics of a social situation. Symptomatic for social isolation is the absence of social relationships and the lack of embeddedness into a community. However, loneliness is the personal experience of a social situation, which can deviate from the objective characteristic. Often definitions of social isolation center on the absence of social relationships. An additional risk lies in the occurrence of poor-quality relationships or negative social interactions. As part of your thesis, you have the opportunity to empirically investigate the relationship between negative social networks and loneliness or to prepare a systematic review.
Agent-Based Modeling
Multidimensional segregation
The model of segregation by Thomas Schelling is a classic, but we hardly know how having more than one attribute affects segregation dynamics. This study uses an agent-based model that I already built, which investigates levels of segregation when agents have more than one attribute and make choices based on different rules (e.g. always make decisions based on one attribute vs. randomly selecting the attribute that matters for making a choice). Interestingly, having individuals care about different dimensions leads to the counterintuitive result that overall more segregation comes about.
Luck vs talent: Why are the luckiest people more likely to be moderately talented?
A recent paper by Pulchino et al. (2018 presents a very simple agent-based model to show the importance of luck vs talent. Their model gives a solution for the puzzle that 1) talent is normally distributed, but 2) success is power-law distributed as discussed in the lecture. Their idea is that while talented people might be more likely to make something out of luck events that present themselves randomly to them, there are just more medium talented people around to begin with (because of the normal distribution of talent). Hence, the luckiest person is more likely to be medium talented. A thesis project could replicate their model and test certain assumptions of it, e.g. how do the dynamics look like when luck does not double capital (as in the paper) but multiplies capital with a different factor. What if instead of multiplication, a luck event adds capital? You could then move on and extend the model further, e.g. what if propensity to turn a luck event into success is not just correlated to talent, but also to the talent/success of other people in the social environment (network or neighborhood)?
Other
Social status and life expectancy: Do you live longer when you win a medal at the Olympics?
Previous research connects social status with life expectancy and shows that actors who win an academy award tend to live longer than those you are nominated, but don’t win the award (Redelmeier and Singh 2001) . This project investigates this phenomenon in the context of Olympic medalists. It collects historical data on sports men/women, their rank in the Olympic final of their discipline and their life expectancy. It applies a regression discontinuity design by comparing the life expectancy of those finalists who won the bronze medal with those finalists who became fourth in the final.
Broken window theory: Do individual level attributes moderate the spreading of disorder?
Generally, I would be very keen to supervise projects in the area of "broken window theory" or the spreading of disorder. I think it is one of the coolest topics ever. One potential project could aim to replicate some of the experiments conducted in Keizer et al. (2008). For example, one could think about attaching flyers to the bikes on campus and have as a treatment a rubbish bag next to the bikes (as in the paper of Keuschnigg and Wolbring 2015). The theory would suggest that more flyers land on the ground when there is a rubbish bag next to the bikes. One extension could be that instead of just counting the flyers, you also classify the bikes according to how "expensive" or "high quality" they seem to be and give them flyers with different numbers on them. You can then find out if students with low quality bikes are more or less likely to throw away the flyers. One major shortcoming in the work of Keizer et al. (2008) is that it ignores characteristics of the individuals. The paper by Keuschnigg and Wolbring (2015) tries to get there by conducting the experiments in different parts of the city with different social capital. Even though this goes into the right direction one can still criticize it for applying higher level measure (social capital of a neighborhood) to a lower level outcome (individuals throwing down flyers).
Say “ cheeeeeeze”! The emergence of smiling faces in advertisements
While we nowadays often smile when being photographed, portraits from the past often show people in a very serious and non-smiling pose. When did people start to smile? And who smiled first? This project draws on 116.746 identified faces (bounding box location on image, predicted age and gender) for all historical advertisements from all 8,840 issues of The Economist magazine, years 1843 to 2014. Data are available here: https://tinyurl.com/4k8s4cu9 . The project then uses machine learning tools, such as OpenCV (see here https://flothesof.github.io/smile-recognition.html ) to detect whether people are smiling or not and relates this information back to dates and demographic characteristics.