Artificial Intelligence (AI)-based systems are widely employed nowadays to make decisions that have far-reaching impact on individuals and society. Their decisions might affect everyone, everywhere, and anytime, entailing concerns about potential human rights issues. Therefore, it is necessary to move beyond traditional AI algorithms optimized for predictive performance and embed ethical and legal principles in their design, training, and deployment to ensure social good while still benefiting from the huge potential of the AI technology. The goal of this survey is to provide a broad multidisciplinary overview of the area of bias in AI systems, focusing on technical challenges and solutions as well as to suggest new research directions towards approaches well-grounded in a legal frame. In this survey, we focus on data-driven AI, as a large part of AI is powered nowadays by (big) data and powerful machine learning algorithms. If otherwise not specified, we use the general term bias to describe problems related to the gathering or processing of data that might result in prejudiced decisions on the bases of demographic features such as race, sex, and so forth.This article is categorized under:

Homophily can put minority groups at a disadvantage by restricting their ability to establish links with a majority group or to access novel information. Here, we show how this phenomenon can influence the ranking of minorities in examples of real-world networks with various levels of heterophily and homophily ranging from sexual contacts, dating contacts, scientific collaborations, and scientific citations. We devise a social network model with tunable homophily and group sizes, and demonstrate how the degree ranking of nodes from the minority group in a network is a function of (i) relative group sizes and (ii) the presence or absence of homophilic behaviour. We provide analytical insights on how the ranking of the minority can be improved to ensure the representativeness of the group and correct for potential biases. Our work presents a foundation for assessing the impact of homophilic and heterophilic behaviour on minorities in social networks.

People's perceptions about the frequency of attributes in their social networks sometimes show false consensus, or overestimation of the frequency of own attributes, and sometimes false uniqueness, or underestimation of the frequency. Here we show that both perception biases can emerge solely from the structural properties of social networks. Using a generative network model, we show analytically that perception biases depend on the level of homophily and its asymmetric nature, as well as on the size of minority group. Model predictions correspond to empirical data from a cross-cultural survey study and to numerical calculations on six real-world networks. We also show in what circumstances individuals can reduce their biases by relying on perceptions of their neighbors. This study advances our understanding of the impact of network structure on perception biases and offers a quantitative approach for addressing related issues in society. arXiv:1710.08601v4 [physics.soc-ph] 22 Jul 2019 Empirical networksThe first network, Brazilian network, captures sexual contact between sex workers and sex buyers 46 . The network consists of 16, 730 nodes and 39, 044 edges. There are 10, 106 sex workers and 6, 624 sex buyers (minority-group size f a = 0.4). In this network, no edges among members of the same group exist resulting in the Newman's assortativity (q = −1), and consequently, the network is purely heterophilic (h = 0).The second network is an online Swedish dating network from PussOKram.com (POK) 47 . This network contains 29, 341 nodes with strong heterophily (h = 0.17, q = −0.65). Given the high bipartivity of the network, we are able to infer the group of nodes using the max-cut greedy algorithm. The results are in good agreement with the bipartivity reported 48 . Since the group definition is arbitrary, we label the nodes based on their relative group size as minority gender and majority gender. Here, the fraction of the minority in the network is 0.44.The third network is a Facebook network of a university in the United States (USF51) 49 . The network is composed of 6,253 nodes and includes information about individuals' gender. In this network male students are in the minority, occupying 42% of the network, and the network exhibits a small heterophily 49 (q = −0.06, h = 0.48).The fourth network is extracted from the collaborative programming environment GitHub. The network is a snapshot of the community (extracted August 4, 2015) that includes information about the first name and family name of the programmers. We used the first name and family name to infer the gender of the programmers 50 . After we removed ambiguous names, the network consisted of 120, 338 men and 7, 330 women. Here, women belong to the minority group and represent only about 6% of the population. The network displays a moderately symmetric gender homophily of 0.53 (q = 0.07).The fifth network depicts scientific collaborations in computer science and is extracted from Digital Bibliography & Library Project's website (DBLP) 51 . We used a new...

Scienti¯c collaborations shape ideas as well as innovations and are both the substrate for, and the outcome of, academic careers. Recent studies show that gender inequality is still present in many scienti¯c practices ranging from hiring to peer-review processes and grant applications. In this work, we investigate gender-speci¯c di®erences in collaboration patterns of more than one million computer scientists over the course of 47 years. We explore how these patterns change over years and career ages and how they impact scienti¯c success. Our results highlight that successful male and female scientists reveal the same collaboration patterns: compared to scientists in the same career age, they tend to collaborate with more colleagues than other scientists, seek innovations as brokers and establish longer-lasting and more repetitive collaborations. However, women are on average less likely to adopt the collaboration patterns that are related with success, more likely to embed into ego networks devoid of structural holes, and they exhibit stronger gender homophily as well as a consistently higher dropout rate than men in all career ages.

Threshold models try to explain the consequences of social influence like the spread of fads and opinions. Along with models of epidemics, they constitute a major theoretical framework of social spreading processes. In threshold models on static networks, an individual changes her state if a certain fraction of her neighbors has done the same. When there are strong correlations in the temporal aspects of contact patterns, it is useful to represent the system as a temporal network. In such a system, not only contacts but also the time of the contacts are represented explicitly. There is a consensus that bursty temporal patterns slow down disease spreading. However, as we will see, this is not a universal truth for threshold models. In this work, we propose an extension of Watts' classic threshold model to temporal networks. We do this by assuming that an agent is influenced by contacts which lie a certain time into the past. I.e., the individuals are affected by contacts within a time window. In addition to thresholds as the fraction of contacts, we also investigate the number of contacts within the time window as a basis for influence. To elucidate the model's behavior, we run the model on real and randomized empirical contact datasets.

No abstract

In this paper, we study the network of global interconnections between language communities, based on shared co-editing interests of Wikipedia editors, and show that although English is discussed as a potential lingua franca of the digital space, its domination disappears in the network of co-editing similarities, and instead local connections come to the forefront. Out of the hypotheses we explored, bilingualism, linguistic similarity of languages, and shared religion provide the best explanations for the similarity of interests between cultural communities. Population attraction and geographical proximity are also significant, but much weaker factors bringing communities together. In addition, we present an approach that allows for extracting significant cultural borders from editing activity of Wikipedia users, and comparing a set of hypotheses about the social mechanisms generating these borders. Our study sheds light on how culture is reflected in the collective process of archiving knowledge on Wikipedia, and demonstrates that cross-lingual interconnections on Wikipedia are not dominated by one powerful language. Our findings also raise some important policy questions for the Wikimedia Foundation.

Sampling from large networks represents a fundamental challenge for social network research. In this paper, we explore the sensitivity of different sampling techniques (node sampling, edge sampling, random walk sampling, and snowball sampling) on social networks with attributes. We consider the special case of networks (i) where we have one attribute with two values (e.g., male and female in the case of gender), (ii) where the size of the two groups is unequal (e.g., a male majority and a female minority), and (iii) where nodes with the same or different attribute value attract or repel each other (i.e., homophilic or heterophilic behavior). We evaluate the different sampling techniques with respect to conserving the position of nodes and the visibility of groups in such networks. Experiments are conducted both on synthetic and empirical social networks. Our results provide evidence that different network sampling techniques are highly sensitive with regard to capturing the expected centrality of nodes, and that their accuracy depends on relative group size differences and on the level of homophily that can be observed in the network. We conclude that uninformed sampling from social networks with attributes thus can significantly impair the ability of researchers to draw valid conclusions about the centrality of nodes and the visibility or invisibility of groups in social networks.Comment: Published at WWW'1