Auditing Algorithms: Understanding Algorithmic Systems from the Outside In

Metaxa, Danaë; Park, Joon Sung; Robertson, Ronald E.; Karahalios, Karrie; Wilson, Christo; Hancock, Jeff; Sandvig, Christian

doi:10.1561/1100000083

Cited by 64 publications

(47 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Other work has focused on organisational challenges and barriers that practitioners face when attempting to build more responsible AI products and services [93,94,126], as well as considerations regarding fairness perceptions across cultures [128]. Research has also been directed specifically towards better understanding AI practitioners' needs and the development of frameworks, processes and tools to help assess and audit algorithmic systems for unfair, biased, or otherwise harmful behaviour (e.g., [14,18,94,99,125]) -both internally (within the organisations responsible for developing and maintaining these systems) and externally (by independent auditors, users and/or regulators). Particularly relevant here is the work exploring the issues and efficacy of fairness-specific tooling for supporting practitioners, for example, that of Holstein et al [65], Lee et al [88] and Deng et al [39], which considered the perceptions and use of so-called "fairness toolkits" that aim to support ML practitioners with fairness concerns, finding significant disconnects between the tooling and the expectations, needs and practices of practitioners.…”

Section: Algorithmic Fairnessmentioning

confidence: 99%

Out of Context: Investigating the Bias and Fairness Concerns of "Artificial Intelligence as a Service"

Lewicki,

Lee,

Cobbe

et al. 2023

Preprint

View full text Add to dashboard Cite

AI as a Service" (AIaaS) is a rapidly growing market, offering various plug-and-play AI services and tools. AIaaS enables its customers (users)-who may lack the expertise, data, and/or resources to develop their own systems-to easily build and integrate AI capabilities into their applications. Yet, it is known that AI systems can encapsulate biases and inequalities that can have societal impact. This paper argues that the context-sensitive nature of fairness is often incompatible with AIaaS' "one-size-fits-all" approach, leading to issues and tensions. Specifically, we review and systematise the AIaaS space by proposing a taxonomy of AI services based on the levels of autonomy afforded to the user. We then critically examine the different categories of AIaaS, outlining how these services can lead to biases or be otherwise harmful in the context of end-user applications. In doing so, we seek to draw research attention to the challenges of this emerging area. CCS CONCEPTS• Computing methodologies → Artificial intelligence; • Social and professional topics → Computing / technology policy; Socio-technical systems.

show abstract

Section: Algorithmic Fairnessmentioning

confidence: 99%

Out of Context: Investigating the Bias and Fairness Concerns of "Artificial Intelligence as a Service"

Lewicki,

Lee,

Cobbe

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…It is now the norm for most platforms to use personalized algorithms to recommend content to their users (Bhandari and Bimo, 2022). Algorithm audits (Metaxa et al, 2021) and internal documents (Wells et al, 2021) indicate that feeds like the TikTok For You page and the Instagram Explore page curate content by identifying users' interests through an analysis of digital trace data, like the posts they look at, like, share, or skip. While the specifics of these processes are rarely made known to the general public-much less everyday users-knowing how these algorithm-based systems work can help people have more agency over what they see and share.…”

Section: Social Media Literacymentioning

confidence: 99%

To use or be used? The role of agency in social media use and well-being

Lee

Ellison

Hancock

2023

Front. Comput. Sci.

View full text Add to dashboard Cite

In this paper, we develop the concept of agentic social media use: a way of engaging with social media that emphasizes having the beliefs, knowledge, and practices to use it intentionally. In comparison to instances of “mindless” social media use, people who use social media agentically do so with a purpose in mind: they leverage the affordances of social media to do things that are meaningful, useful, or satisfying for them. For example, people can use social media to intentionally build or manage their relationships, to seek out and learn new information about their interests, or to craft a positive image of themselves through the content they post. Crucially, however, there are many other valuable uses of social media that may not be considered conventionally productive but are nonetheless deliberate and useful, such as using social media intentionally to relax, unwind, and entertain themselves in an effort to modulate their emotions. To use social media agentically means to (1) hold an agentic mindset about one's relationship with social media, (2) have the knowledge and literacy to understand how to navigate social media effectively, and (3) enact practices that assert control over specific elements of social media use, such as curating content and refining algorithmic recommendation. Approaching social media use from the perspective of agency and intentionality allows us to better understand heterogeneous social media effects and to identify new ways of helping people benefit from these technologies.

show abstract

“…Past research has successfully identied harmful and discriminatory behaviors in a variety of algorithmic domains, including search engines [52,54,60], online advertising [45,67], facial recognition [10], word embedding [7], and e-commerce [40]. For example, Buolamwini and Gebru audited three commercial gender classication systems and found that the commercial systems misclassied darker-skinned women more often than white people [10].…”

Section: Algorithm Auditmentioning

confidence: 99%

“…The presence of biases in algorithmic systems has given rise to auditing approaches, usually led by AI/ML experts, to investigate these systems for harmful behaviors [17,52]. These expert-driven auditing techniques have been successful in nding and mitigating many cases of harmful algorithmic behavior; yet, they suer from a number of limitations.…”

Section: Introductionmentioning

confidence: 99%

Participation and Division of Labor in User-Driven Algorithm Audits: How Do Everyday Users Work together to Surface Algorithmic Harms?

Kingsley

Fan

et al. 2023

Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems

View full text Add to dashboard Cite

Recent years have witnessed an interesting phenomenon in which users come together to interrogate potentially harmful algorithmic behaviors they encounter in their everyday lives. Researchers have started to develop theoretical and empirical understandings of these user-driven audits, with a hope to harness the power of users in detecting harmful machine behaviors. However, little is known about users' participation and their division of labor in these audits, which are essential to support these collective eorts in the future. Through collecting and analyzing 17,984 tweets from four recent cases of user-driven audits, we shed light on patterns of users' participation and engagement, especially with the top contributors in each case. We also identied the various roles users' generated content played in these audits, including hypothesizing, data collection, amplication, contextualization, and escalation. We discuss implications for designing tools to support user-driven audits and users who labor to raise awareness of algorithm bias. CCS CONCEPTS• Human-centered computing ! Human computer interaction (HCI); Empirical studies in HCI .

show abstract

Auditing Algorithms: Understanding Algorithmic Systems from the Outside In

Cited by 64 publications

References 0 publications

Out of Context: Investigating the Bias and Fairness Concerns of "Artificial Intelligence as a Service"

Out of Context: Investigating the Bias and Fairness Concerns of "Artificial Intelligence as a Service"

To use or be used? The role of agency in social media use and well-being

Participation and Division of Labor in User-Driven Algorithm Audits: How Do Everyday Users Work together to Surface Algorithmic Harms?

Contact Info

Product

Resources

About