Team: Rua M. Williams, Jeremy Waisome, Kyla McMullen, Emma Drobina

This past week,  I co-presented a workshop at the 2019 ACM Tapia Celebration of Diversity in Computing. Inspired by Hecht et al.’s proposal to change the peer-review process to account for potential societal impacts, we guided attendees through evaluations of the potential positive and negative affects of  a series of case studies. In a world where computers are increasingly intertwined with human lives, we cannot simply say that computers are objective. Instead, we as researchers and developers must understand that machines are only as objective as the people who create and deploy them.

The workshop slides and a transcript of their content can be viewed below.

  • The Fallacy of Objectivity: Leveraging Situated Knowledges to Advance Computing for Social Justice
    • Rua M. Williams/@starfeuri: Headshot of Rua M. Williams, a white nonbinary person with short blonde hair and blue eyes in a blue button up shirt and brown tie.
    • Jeremy Waisome/@jeremywaisome: Photo of Dr. Jeremy Waisome, a Black, female, postdoctoral associate with an afro ponytail and an energetic smile. She is wearing a blue and white striped button-down with a white collar. It looks better than it sounds.
    • Kyla McMullen/@Dr_Kyla: Headshot of Kyla McMullen, a Black woman with braids in her hair, wearing a checkered suit jacket and white shirt.
    • Emma Drobina/@emmadrobina: Headshot of Emma Drobina, a white person with short dark hair wearing glasses and a green flannel shirt.
  • What do we mean by “The Fallacy of Objectivity”?
    • “Objectivity” as theorized by modern science requires removing one’s self from one’s own observations. This feat is often regarded as impossible for marginalized researchers. Their subjectivity is always present. The only researcher afforded the privilege of self-invisibility is the hegemonic body. As marginalized people, we know the hegemony is only invisible to itself.
    • This isn’t your grandmother’s bias talk. The discourse around implicit/explicit bias has led us to perceive bias as something measurable and containable that can be contained and controlled. Bias-talk leads us to ad-hoc solutions. Understanding Objectivity as a Lie leads us to reshape our very practice of research inquiry.
  • Hecht et al. 2018 “It’s time to do something…”
    • Researchers should explicitly evaluate potential positive and negative impacts of their work in their publications
    • Reviewers should consider these impacts, question authors if they are missing, and evaluate the work based on societal factors
    • Some example scenarios within this paper fail to adequately address societal impacts from the perspective of marginalized user groups
      • “Consider a researcher who develops a new access technology that supports people with disabilities in independently doing something that used to require human help. This technology may dramatically decrease the cost of providing this support and make the service more easily available to people who need it. However, if the services were previously provided by an employee, the new technology may lead to fewer jobs. Also, this technology likely does not have a social component, which users of the service may have highly valued, but which is now unavailable to them. The researcher who develops this technology would be required under our recommendations to enumerate not only the benefits to accessibility, but also the negative impacts on employment and social interaction. The researcher would also likely choose to highlight policies or new technologies that could mitigate these negative impacts (e.g. the integration of remote social support technologies).”
      • Problem: This example objectifies the disabled person as a site of employment for an able-bodied carer rather than as an autonomous user of technology. The negative impacts to society are framed as economic impacts to an exclusively able society and social connection as charity. Though such a technology may allow the disabled user to pursue their own employment or mutual social engagements, this impact is somehow inconceivable to the authors.
  • Scenario E:
    • Imagine an educational app created to teach children about history. To make it engaging and interactive, it is designed as a game – as players progress through the lessons, they will collect items and gain points.
      • Positive: Could increase student engagement with the app and make learning about history more fun.
      • Negatives: Is this a historical topic where gamification is appropriate? Making the app a game could be seen as making light of serious historical topics (e.g. slavery).
  • Scenario R:
    • Imagine a risk assessment and cost/benefit analysis system designed to support insurance underwriters in determining coverage and necessity of medications and procedures prescribed by hospital staff in in-patient settings. This system uses demographic features for prediction such as race, sex, residence, immigrant status, disability, and health condition.
      • Positive: potentially more cost-effective, efficient, and more efficacious treatment delivered to patients in dire need
      • Negative: Insurance provider control over what is “medically necessary” can endanger lives, lead to long term complications, and death.
  • Scenario K:
    • Imagine a 3D audio system (a system whereby a listener hears audio as though it is being emitted from a specified point in space) that renders audio, based on the “average human physiological proportions”. The 3D audio rendering system is used in most commercial virtual audio applications”.
      • Positive: positional audio for people who need it. No need for costly measurement. Great application areas.
      • Negative: the KEMAR or “standard person” used was only measured on White men and tests for accuracy are only performed using audiophiles (people who are trained listeners. The “standard person” does not take into account physiological differences between people. Accuracy is defined by expert performance, rather than an “average listener’s” performance
  • Scenario J:
    • Envision a 10-week summer research program designed to prepare late-stage undergraduate students for careers in research, including professional development to enhance students understanding of graduate school. A specific aim of the program increase minority student participation in computing research.
      • Positive: Students from underrepresented backgrounds have the opportunity to participate in summer research
      • Negative: How is it framed? It’s important to keep in mind that there is a difference between being a minority vs. being minoritized.
  • Other Scenarios for Homework #WhoseObjectivity
    • Imagine a system that detects what users on social media are posting about and recommends them other users that share their interests.
    • Imagine a system that monitors social media content and is designed to flag users who are expressing suicidal thoughts or acts and alert authorities.
    • Imagine an app that reports crime in real time so that users can avoid certain areas of town.
  • Homework: HARD MODE
    • Put your own past work up as a scenario for others to evaluate.