An Introduction to Identifying and Mitigating Online Bias
A series of research perspectives in human-computer interaction.
The online availability of digital information, social networking, and automated computing has become indispensable for daily life and particularly essential to the recent relocation of jobs, education, entertainment, and social functions from physical spaces to the web. This change has occurred alongside the increase in COVID-19 infection rates, filled hospital beds, and protective measures against the pandemic including the buffer of device screens that have mediated social connection and information about the world outside. However, as much as the digital world might have become a pacifier or a temporary solution for reality, it is not immune to its influence.
Social issues have grown severe due to the tenacity of disproportionate norms and values, and the optimism in our 2020 vision quickly revealed a clear picture of a rather distorted world. As more of our activity and decisions continue to occur online, it is often difficult or impossible to estimate biased perception in this environment. Although slight at first, when bias continues to propagate unchecked, the continued reliance on the internet can lead to discrimination or unethicality experienced by an individual or society. Bias influences individual mental models, social interactions, and real-world decisions; and like some of the largest issues in society, they are found systemically in the online environment.
This series aims to explore the way bias distorts people’s perception of what is normal and acceptable in their use of online applications. In particular, it attempts to bring attention to the subject of bias as it relates to perspectives in human-computer interaction (HCI). Disciplines ranging from psychology, sociology, communications, to economy have investigated the presence and effects of bias on behavior and society, and theoretical insights into the subject are better understood by research in these areas. Instead, I will focus on the kinds of bias that have emerged specifically from online activity and explore their effects.
In the context of cognitive science, psychology, and behavioral economics; bias is a systematic deviation from a norm or from rational judgement. In a statistical sense, it is a systematic deviation from a normal or prior distribution. Bias may be behavioral, such as in a particular perception, assumption, or judgment; or it may be cultural and exist at a social, economic, political, or technological level. The definitions of cognitive bias vary, as well as the methods used to measure it in psychology. The Implicit Association Test (IAT) is a common assessment, which measures the strength of subconscious associations in the mental representations of concepts including real-world objects and people as well as social constructs [1]. The IAT is available online at Project Implicit, where you can test yourself on a variety of topics.
The widespread relevance of the subject reflects the multitude of ways it exists in human behavior, including in the interaction with technology. The discipline of HCI is well-positioned for the examination of bias that exists and evolves with the growing usage of online tasks and artifacts. In this six-part series, I will explore four areas in which bias pollutes the online environment: information search [2/6], social media [3/6], algorithmic systems [4/6], and online education [5/6]. Of course these areas aren’t mutually exclusive in current online applications, but the rough distinctions serve to describe the unique kinds of bias present in each, their prevalence and impact, and specific methods for their mitigation. This sets the backdrop for following evaluations of bias in applications that merge in these areas.
References
- Greenwald, A., McGhee, D., & Schwartz, J.L. (1998). Measuring individual differences in implicit cognition: the implicit association test. Journal of personality and social psychology, 74(6), 1464–80. https://doi.org/10.1037/0022-3514.74.6.1464
