Search

The increasing reliance on complex algorithmic systems by online platforms has sparked a growing need for algorithm auditing, a research methodology evaluating these systems' functionality and societal impact. In this paper, we systematically review algorithm auditing studies and identify trends in their methodological approaches, the geographic distribution of authors, and the selection of platforms, languages, geographies, and group-based attributes in the focus of auditing research. We present evidence of a significant skew of research focus toward Western contexts, particularly the US, and a disproportionate reliance on English language data. Additionally, our analysis indicates a tendency in algorithm auditing studies to focus on a narrow set of group-based attributes, often operationalized in simplified ways, which might obscure more nuanced aspects of algorithmic bias and discrimination. By conducting this review, we aim to provide a clearer understanding of the current state of the algorithm auditing field and identify gaps that need to be addressed for a more inclusive and representative research landscape.

Internet users highly rely on and trust web search engines, such as Google, to find relevant information online. However, scholars have documented numerous biases and inaccuracies in search outputs. To improve the quality of search results, search engines employ various content moderation practices such as interface elements informing users about potentially dangerous websites and algorithmic mechanisms for downgrading or removing low-quality search results. While the reliance of the public on web search engines and their use of moderation practices is well-established, user attitudes towards these practices have not yet been explored in detail. To address this gap, we first conducted an overview of content moderation practices used by search engines, and then surveyed a representative sample of the US adult population (N=398) to examine the levels of support for different moderation practices applied to potentially misleading and/or potentially offensive content in web search. We also analyzed the relationship between user characteristics and their support for specific moderation practices. We find that the most supported practice is informing users about potentially misleading or offensive content, and the least supported one is the complete removal of search results. More conservative users and users with lower levels of trust in web search results are more likely to be against content moderation in web search.

Prediction-based decision-making systems are becoming increasingly prevalent in various domains. Previous studies have demonstrated that such systems are vulnerable to runaway feedback loops, e.g., when police are repeatedly sent back to the same neighborhoods regardless of the actual rate of criminal activity, which exacerbate existing biases. In practice, the automated decisions have dynamic feedback effects on the system itself that can perpetuate over time, making it difficult for short-sighted design choices to control the system's evolution. While researchers started proposing longer-term solutions to prevent adverse outcomes (such as bias towards certain groups), these interventions largely depend on ad hoc modeling assumptions and a rigorous theoretical understanding of the feedback dynamics in ML-based decision-making systems is currently missing. In this paper, we use the language of dynamical systems theory, a branch of applied mathematics that deals with the analysis of the interconnection of systems with dynamic behaviors, to rigorously classify the different types of feedback loops in the ML-based decision-making pipeline. By reviewing existing scholarly work, we show that this classification covers many examples discussed in the algorithmic fairness community, thereby providing a unifying and principled framework to study feedback loops. By qualitative analysis, and through a simulation example of recommender systems, we show which specific types of ML biases are affected by each type of feedback loop. We find that the existence of feedback loops in the ML-based decision-making pipeline can perpetuate, reinforce, or even reduce ML biases.

Nowadays, many social media platforms are centered around content creators (CC). On these platforms, the tie formation process depends on two factors: (a) the exposure of users to CCs (decided by, e.g., a recommender system), and (b) the following decision-making process of users. Recent research studies underlined the importance of content quality by showing that under exploratory recommendation strategies, the network eventually converges to a state where the higher the quality of the CC, the higher their expected number of followers. In this paper, we extend prior work by (a) looking beyond averages to assess the fairness of the process and (b) investigating the importance of exploratory recommendations for achieving fair outcomes. Using an analytical approach, we show that non-exploratory recommendations converge fast but usually lead to unfair outcomes. Moreover, even with exploration, we are only guaranteed fair outcomes for the highest (and lowest) quality CCs.

Binary decision making classifiers are not fair by default. Fairness requirements are an additional element to the decision making rationale, which is typically driven by maximizing some utility function. In that sense, algorithmic fairness can be formulated as a constrained optimization problem. This paper contributes to the discussion on how to implement fairness, focusing on the fairness concepts of positive predictive value (PPV) parity, false omission rate (FOR) parity, and sufficiency (which combines the former two). We show that group-specific threshold rules are optimal for PPV parity and FOR parity, similar to well-known results for other group fairness criteria. However, depending on the underlying population distributions and the utility function, we find that sometimes an upper-bound threshold rule for one group is optimal: utility maximization under PPV parity (or FOR parity) might thus lead to selecting the individuals with the smallest utility for one group, instead of selecting the most promising individuals. This result is counter-intuitive and in contrast to the analogous solutions for statistical parity and equality of opportunity. We also provide a solution for the optimal decision rules satisfying the fairness constraint sufficiency. We show that more complex decision rules are required and that this leads to within-group unfairness for all but one of the groups. We illustrate our findings based on simulated and real data., Comment: 14 pages, accepted to ACM FAccT 2022, code available on GitHub: https://github.com/joebaumann/fair-prediction-based-decision-making

Parler, a "free speech" platform popular with conservatives, was taken offline in January 2021 due to the lack of moderation of hateful and QAnon- and other conspiracy-related content that allegedly made it instrumental in the organisation of the storming of the US Capitol on January 6. However, Parler co-existed with other social media platforms, and comparative studies are needed to draw conclusions about the prevalence of anti-social language, hate speech, or conspiracy theory content on the platform. We address this through a cross-platform comparison of posts related to QAnon. We compare posts with the hashtag #QAnon on Parler over a month-long period with posts on Twitter and Gab. In our analysis, Parler emerges as the platform with the highest average toxicity of the posts, though this largely stems from the distinctive way hashtags are used on this platform. Gab has the highest proportion of #QAnon posts with hate terms, and Parler and Twitter are similar in this respect. On all three platforms, posts mentioning female political figures, Democrats, or Donald Trump have more anti-social language than posts mentioning male politicians, Republicans, or Joe Biden. An analysis of entities mentioned in posts revealed differences in content - with Twitter posts mentioning prominent figures related to QAnon, while Parler and Gab posts mention entities related to conspiracies. Narrative analysis indicates that the discussion on Twitter centres on QAnon and is critical of it, Parler focuses on supporting Donald Trump, while on Gab the discussion focuses on more conspiratorial content, in relation to both Trump and other political figures.

Two-sided matching markets have long existed to pair agents in the absence of regulated exchanges. A common example is school choice, where a matching mechanism uses student and school preferences to assign students to schools. In such settings, forming preferences is both difficult and critical. Prior work has suggested various prediction mechanisms that help agents make decisions about their preferences. Although often deployed together, these matching and prediction mechanisms are almost always analyzed separately. The present work shows that at the intersection of the two lies a previously unexplored type of strategic behavior: agents returning to the market (e.g., schools) can attack future predictions by interacting short-term non-optimally with their matches. Here, we first introduce this type of strategic behavior, which we call an `adversarial interaction attack'. Next, we construct a formal economic model that captures the feedback loop between prediction mechanisms designed to assist agents and the matching mechanism used to pair them. This economic model allows us to analyze adversarial interaction attacks. Finally, using school choice as an example, we build a simulation to show that, as the trust in and accuracy of predictions increases, schools gain progressively more by initiating an adversarial interaction attack. We also show that this attack increases inequality in the student population.

Perhaps the most controversial questions in the study of online platforms today surround the extent to which platforms can intervene to reduce the societal ills perpetrated on them. Up for debate is whether there exist any effective and lasting interventions a platform can adopt to address, e.g., online bullying, or if other, more far-reaching change is necessary to address such problems. Empirical work is critical to addressing such questions. But it is also challenging, because it is time-consuming, expensive, and sometimes limited to the questions companies are willing to ask. To help focus and inform this empirical work, we here propose an agent-based modeling (ABM) approach. As an application, we analyze the impact of a set of interventions on a simulated online dating platform on the lack of long-term interracial relationships in an artificial society. In the real world, a lack of interracial relationships are a critical vehicle through which inequality is maintained. Our work shows that many previously hypothesized interventions online dating platforms could take to increase the number of interracial relationships from their website have limited effects, and that the effectiveness of any intervention is subject to assumptions about sociocultural structure. Further, interventions that are effective in increasing diversity in long-term relationships are at odds with platforms' profit-oriented goals. At a general level, the present work shows the value of using an ABM approach to help understand the potential effects and side effects of different interventions that a platform could take.

Online knowledge platforms such as Stack Overflow and Wikipedia rely on a large and diverse contributor community. Despite efforts to facilitate onboarding of new users, relatively few users become core contributors, suggesting the existence of barriers or hurdles that hinder full involvement in the community. This paper investigates such issues on Stack Overflow, a widely popular question and answer community for computer programming. We document evidence of a "leaky pipeline", specifically that there are many active users on the platform who never post an answer. Using this as a starting point, we investigate potential factors that can be linked to the transition of new contributors from asking questions to posting answers. We find a user's individual features, such as their tenure, gender, and geographic location, as well as features of the subcommunity in which they are most active, such as its size and the prevalence of negative social feedback, have a significant relationship with their likelihood to post answers. By measuring and modeling these relationships our paper presents a first look at the challenges and obstacles to user promotion along the pipeline of contributions in online communities.

Academic fields exhibit substantial levels of gender segregation. To date, most attempts to explain this persistent global phenomenon have relied on limited cross-sections of data from specific countries, fields, or career stages. Here we used a global longitudinal dataset assembled from profiles on ORCID.org to investigate which characteristics of a field predict gender differences among the academics who leave and join that field. Only two field characteristics consistently predicted such differences: (1) the extent to which a field values raw intellectual talent ("brilliance") and (2) whether a field is in Science, Technology, Engineering, and Mathematics (STEM). Women more than men moved away from brilliance-oriented and STEM fields, and men more than women moved toward these fields. Our findings suggest that stereotypes associating brilliance and other STEM-relevant traits with men more than women play a key role in maintaining gender segregation across academia., Comment: 28 pages of combined manuscript and supplemental materials

Despite the potential of online sharing economy platforms such as Uber, Lyft, or Foodora to democratize the labor market, these services are often accused of fostering unfair working conditions and low wages. These problems have been recognized by researchers and regulators but the size and complexity of these socio-technical systems, combined with the lack of transparency about algorithmic practices, makes it difficult to understand system dynamics and large-scale behavior. This paper combines approaches from complex systems and algorithmic fairness to investigate the effect of algorithm design decisions on wage inequality in ride-hailing markets. We first present a computational model that includes conditions about locations of drivers and passengers, traffic, the layout of the city, and the algorithm that matches requests with drivers. We calibrate the model with parameters derived from empirical data. Our simulations show that small changes in the system parameters can cause large deviations in the income distributions of drivers, leading to a highly unpredictable system which often distributes vastly different incomes to identically performing drivers. As suggested by recent studies about feedback loops in algorithmic systems, these initial income differences can result in enforced and long-term wage gaps., Comment: Code for the simulation can be found at https://github.com/bokae/taxi

Programming is a valuable skill in the labor market, making the underrepresentation of women in computing an increasingly important issue. Online question and answer platforms serve a dual purpose in this field: they form a body of knowledge useful as a reference and learning tool, and they provide opportunities for individuals to demonstrate credible, verifiable expertise. Issues, such as male-oriented site design or overrepresentation of men among the site's elite may therefore compound the issue of women's underrepresentation in IT. In this paper we audit the differences in behavior and outcomes between men and women on Stack Overflow, the most popular of these Q&A sites. We observe significant differences in how men and women participate in the platform and how successful they are. For example, the average woman has roughly half of the reputation points, the primary measure of success on the site, of the average man. Using an Oaxaca-Blinder decomposition, an econometric technique commonly applied to analyze differences in wages between groups, we find that most of the gap in success between men and women can be explained by differences in their activity on the site and differences in how these activities are rewarded. Specifically, 1) men give more answers than women and 2) are rewarded more for their answers on average, even when controlling for possible confounders such as tenure or buy-in to the site. Women ask more questions and gain more reward per question. We conclude with a hypothetical redesign of the site's scoring system based on these behavioral differences, cutting the reputation gap in half.

Novelty is a key ingredient of innovation but quantifying it is difficult. This is especially true for visual work like graphic design. Using designs shared on an online social network of professional digital designers, we measure visual novelty using statistical learning methods to compare an images features with those of images that have been created before. We then relate social network position to the novelty of the designers images. We find that on this professional platform, users with dense local networks tend to produce more novel but generally less successful images, with important exceptions. Namely, users making novel images while embedded in cohesive local networks are more successful., Comment: accepted to 10th International ACM Web Science Conference, 2018, May 27-30, Amsterdam, The Netherlands, 11 pages, 6 figures, 60 references

Web search is an integral part of our daily lives. Recently, there has been a trend of personalization in Web search, where different users receive different results for the same search query. The increasing level of personalization is leading to concerns about Filter Bubble effects, where certain users are simply unable to access information that the search engines' algorithm decides is irrelevant. Despite these concerns, there has been little quantification of the extent of personalization in Web search today, or the user attributes that cause it. In light of this situation, we make three contributions. First, we develop a methodology for measuring personalization in Web search results. While conceptually simple, there are numerous details that our methodology must handle in order to accurately attribute differences in search results to personalization. Second, we apply our methodology to 200 users on Google Web Search and 100 users on Bing. We find that, on average, 11.7% of results show differences due to personalization on Google, while 15.8% of results are personalized on Bing, but that this varies widely by search query and by result ranking. Third, we investigate the user features used to personalize on Google Web Search and Bing. Surprisingly, we only find measurable personalization as a result of searching with a logged in account and the IP address of the searching user. Our results are a first step towards understanding the extent and effects of personalization on Web search engines today.

Online platforms are an increasingly popular tool for people to produce, promote or sell their work. However recent studies indicate that social disparities and biases present in the real world might transfer to online platforms and could be exacerbated by seemingly harmless design choices on the site (e.g., recommendation systems or publicly visible success measures). In this paper we analyze an exclusive online community of teams of design professionals called Dribbble and investigate apparent differences in outcomes by gender. Overall, we find that men produce more work, and are able to show it to a larger audience thus receiving more likes. Some of this effect can be explained by the fact that women have different skills and design different images. Most importantly however, women and men position themselves differently in the Dribbble community. Our investigation of users' position in the social network shows that women have more clustered and gender homophilous following relations, which leads them to have smaller and more closely knit social networks. Overall, our study demonstrates that looking behind the apparent patterns of gender inequalities in online markets with the help of social networks and product differentiation helps us to better understand gender differences in success and failure., Comment: in The International AAAI Conference on Web and Social Media (ICWSM2017), Montreal, May 2017