AI is “expected to mimic, augment, or displace human agency” (Joyce et al 2021, 2)

• Social factors influence the kinds of questions that get asked, and what constitutes an ‘appropriate’ answer
  (Martin 1991; Callon 1984; Kuhn 1970)
• Present in data representations, objective functions, analysis of outcome, etc.
  (Sismondo 2009, Ch 6; Amann & Knorr Cetina 1988; Gould 1981)
• “Rendering the world technical”
  Tania Murray Li & Samantha Breslin

• Data and evidence are not ‘neutral’
  (Gould 1981; Amann and Knorr Cetina 1988)
• Political, institutional, and economic forces favor the interests of powerful players (DeepMind, OpenAI, etc)
  (Sismondo 2009, Ch 17; Wolfe 2018)
• The training, testing, and provision of AI tools requires extensive (and usually exploitative) human labor
  (Roberts 2020)

• AI and ML are used in healthcare, policing, marketing, sentencing, profiling, content moderation, …
• The uses of AI are very flexible, but tend to be used more effectively by those in power
  (Adams 2002)
• AI may be especially compatible with oppressive structures
  (Winner 1980)

• AI is playing an increasingly central role in everyday life (social media, news articles, navigation, …)
• Lay people incorporate strategies and knowledge into their interactions with AI (e.g. curation)
  (Callon 1984)
• Culture is increasingly performed by AI through bots, algorithmic curation, and creative output
  (Benjamin 2019; Joyce et al. 2021)