Seminars: 2017 Fall


2017 Fall


ANIKÓ HANNÁK (CEU CNS)New Faces of Bias in Online Labor Markets

The internet is fundamentally changing the labor economy. Millions of people use sites like LinkedIn, Upwork or Dribbble to find employment. These services are often driven by algorithms that rate, sort, recommend, and match workers and employers. In theory, many of the mechanisms that cause discrimination in traditional labor markets - cognitive bias, network homophily, statistical discrimination - should be absent from online markets. However, recent studies indicate that these mechanisms do transfer to online platforms, where they may be exacerbated by seemingly harmless design choices. 
In this talk I will investigate three techniques that online platforms use to match users with content: social network algorithms, search algorithms and public review systems. Specifically, I present case studies of 6 different employment platforms, using large scale user data from the employers perspective. I show that biases known from traditional labor markets are indeed present in online platforms, although they manifest in new ways. First, I present results that focus on the visibility of users, which directly impacts the chances of being selected for a job or selling a product. I find that women often receive lower visibility either due to their ranking in the sites’ search interface, or their positions in the underlying social network. Furthermore, I investigate social feedback and other success measures found on user profiles, another important factor in hiring decisions. Overall, my investigations show that demographic features are often correlated with the attention and the social feedback workers and employees receive. Exploring these new forms of inequalities, understanding where social biases enter systems and which mechanisms reinforce them, can be crucial for developing mitigation strategies.


PÉTER ÉRDI (Kalamazoo College)Ranking: The reality, illusion and manipulation of objectivity

Since we humans (1) love lists; (2), are competitive and (3), are jealous of other people, we like ranking. The practice of ranking is studied in social psychology and political science, the algorithms of ranking in computer science. Are these algorithms reflect real objectivity or its illusion only? “Reputation management” admittedly attempts to modify the ideally objective image. As an illustration, recursive ranking – from web pages to patents – is analyzed with the hope of being able to trace new fields of technology. We all know in this room that the challenging question for the future is how to combine human and machine intelligence... 


SREBRENKA LETINA (CEU CNS)Putting individuals back into the network: Some theoretical and analytical approaches on the individual-actor level attributes’ effects on network structure

In well-balanced social network research, besides focusing on social structure, usually some attributes of social actors are usually taken into account in order to arrive at more sophisticated inferences about social processes in the system. While the effect of actor attributes may be conceptualized through processes of social selection, social influence, or co-evolution of structure and attributes, the nature of those attributes depends on research question, and the methodological approaches used for analyzing them are work in progress.

In this seminar, we will start with presenting an application of auto-logistic actor attribute models in examining the role of some demographic and career actor attributes on the co-authorship networks of social scientists in Croatia. We will proceed with a more general overview of theory and previous research on the role of individual psychological attributes of actors on network structure, and present one project that aims to use experimental design employing the analysis of triads for studying this issue.



The paper focuses on the impact factors of drug use awareness and habits (particularly of new psychoactive substances) of primary school students living in segregated settlements in Hungary. The research included an online data collection with responses from more than 3000 lower secondary students. The results of the quantitative analyses were further investigated during a qualitative research that was conducted in 3-3 schools of Baranya, Borsod-Abaúj-Zemplén and Szabolcs-Szatmár-Bereg counties: 5th and 8th graders, their parents, teachers and members of the child protection network were all subjects of interviews. The result of the multivariate data analyses of the online database shows that besides parenting style, lifestyle and general contentment with life, school has a significant effect on the drug use habits of students. Those who think of themselves as bad students, and doesn’t feel like they belong to a community in school are more exposed to the possibility of drug usage. Based on the qualitative research most of the participating 5th graders had extensive knowledge on different types of drugs, and had acquaintances who already tried drugs, but were not users themselves (unlike in the case of cigarettes or alcohol). Among 8th graders we met with substance users too. Participating schools were not equipped to prepare students how to deal with difficulties (the level of creativity and problem-solving skills was alarmingly low), and it is a common practice among students to use drugs to relieve the tension of everyday life. Our experience was that parents and teachers are less informed about drugs than the students, and doesn’t treat it as a high risk to the children. Means to request or offer help are generally lacking, students can only recall external, power solutions (e.g. calling the police) when asked about how could they help their drug using peer. Beyond the analyses the paper offers potential solutions too.


TIMON ELMER (ETH Zürich)A closer look at mental health in friendship and social interaction networks

There is a strong interplay between someone’s social environment and his/her mental health. These inter- and intrapersonal systems are highly dynamic and affect each other constantly through various social mechanisms. Hence, in my work, I aim at understanding the role of three social mechanisms – social integration, social influence, and social selection – in the co-evolution of social interactions and friendship ties with an individual’s mental health. In my talk, I will present two studies that investigate this interplay. The first one focuses on the co-evolution of weak and strong friendship ties with emotional well-being. The second study examines the social selection effects of depressive symptoms on face-to-face social interactions. 


TANJA SLISKOVIC (University of Zagreb)Sharing Values Matters - The Role of Value Homophily in Asking for Advice Among Employees

In this paper we investigate what influences task advice and information seeking among employees. Beyond employee social network structure there is much to be researched in terms of what kind of relationships enable advice seeking and information flow. We use inferential social network analysis (Exponential Random Graph Models) to look into predictors of task advice among employees in a company. We show that perceived value homophily is a predictor of advice seeking, while real value similarity is not. Previous research has indicated that affective and cognitive trust matter when asking for advice or information from co-workers. Still, the role of shared values has not been extensively examined in this sense even though there is plethora of evidence indicating that (value) homophily predicts social tie formation. Our research adds to theories of organizational learning and knowledge dissemination within organizations.




JOHANNES WACHS (CEU CNS)A Network Perspective on Public Procurement Markets

Public procurement markets are where the state buy goods and services from the private sector, accounting for between 5 and 20% of GDP in most countries. Procurement markets are vulnerable to two main forms of cost-multiplying bad behaviors: corruption, when state actors rig the procurement for selected firms, and collusion, when groups of firms agree to forgo competition and charge higher prices. I argue that a network approach offers new perspectives on both of these issues using data from Hungary, the Czech Republic, and Georgia. The network of contracting relationships between issuers and firms and the network of co-bidding behavior among firms are valuable maps for researchers, and anti-corruption and competition authorities alike.