AI Infrastructure and Ethics

Think

Behind the Terms.

What does AI know about you or think it knows? 

AI tools in education are marketed using polished, powerful terms. 

But what might these terms hide? 

Click each term to uncover what’s behind the language and ask yourself: 

How might these words obscure power or risk? 

These words are powerful, not because they’re precise, but because they sound good. 

Reflect:

1. Which of these terms have you encountered before in school or tech marketing? 

2. What new questions might you now ask when you see them? 

Explore

Pause. Now think.

You close your laptop. The video’s done. The logos, the contracts, the surveillance claims… It’s a lot to take in. You wonder: Is this really what’s behind the tools I use to learn, teach, or support others every day? 

Take a moment to sit with that. 

Now let’s walk through five moments of reflection, each grounded in academic scholarship. 

1. The Illusion of Help 

Shoshana Zuboff (2019) explains that digital platforms are built to extract behavioral data, model your actions, and sell predictions about what you’ll do next. 

AI tools may feel helpful or even empowering, but underneath the smooth interface may be systems that monitor you silently. When the product is “free,” your data is often the cost. 

Ask yourself: 
If your behavior is being harvested to train powerful systems (for education, yes, but also for commerce or surveillance) how free is your learning? 

2. The Same Cloud for School and War 

AI is used extensively by militaries around the world, dramatically changing modern warfare. These technologies enable the lethal autonomous weapon systems (ballistic missiles and drones), that have been deployed in the ongoing conflicts in Gaza, Iran and the Ukraine. AI is also being used for military decision making, from identifying potential targets, to evaluating risk. More broadly, AI can be used to monitor populations through broad surveillance, and to manipulate public opinion by spreading propaganda and misinformation.

These applications of AI are not incidental: for many years, defense agencies and states have been funding AI research with military agendas in mind. For example, DARPA (the R&D arm of American defense) extensively funds both private contractor research and public AI research (including at UBC and U of T). In a recent paper on DARPA funding of AI research, researchers argue that the prevalence of these grants in the context of AI research effectively enlists academic researchers in a warfighting agenda, concluding that “this mutual enlistment is crucial to the perpetuation of the military-industrial-commercial-academic complex, and to the technopolitical imaginaries of security through military domination that keep public funds flowing to projects in more efficient killing and destruction, and away from experiments in creative diplomacy, de-escalation, and alternatives to militarisation as a basis for our collective security”.

The private AI industry is also deeply entangled with the military. Billions of dollars are spent by the American military on contracts with technology companies (including Amazon, Google, Microsoft and Oracle), and to support private AI research and development. Although military applications are unlikely to run on the same platforms or to use the same tools that are used for education, these tools were developed by the same companies, and are based on the same underlying technologies. 

3. Whose Knowledge? Whose Power?

Nick Couldry and Ulises Mejias (2019, 2021) show how the global expansion of data systems echoes historical colonial structures. This “data colonialism” appropriates everyday life, with speech, clicks, and routines being quantified and commodified. 

We might also consider how AI technologies relate to (and often interfere with) Indigenous sovereignty. When Indigenous texts, histories and knowledge are used to train AI networks, are Indigenous peoples consulted in the governance and development of these systems, or is knowledge extracted from Indigenous cultures exploitatively? Indigenous advocacy groups and researchers underscore the importance of respecting Indigenous sovereignty and Indigenous authority over traditional knowledge. 

AI models encode a worldview: whose voices matter, which languages are supported, what kinds of knowledge are valued, and how cultural histories are told. These decisions shape the landscape of learning, invisibly but powerfully. 

Ask yourself: 
What does it mean when your knowledge journey is shaped by tools designed elsewhere, by people you’ve never met, for purposes you didn’t choose? What parts of your own knowledge or culture might be left out? 

4. Bias in the System and in the Signal 

Benjamin, R. (2018) argues that algorithms, especially in search engines, reflect and reinforce structural inequalities. When bias is built into surveillance tools, some students may be flagged more often, because of how the system interprets language, appearance, or behavior, rather than because of actual risk.

For example, a peer-reviewed study by researchers Buolamwini and Gebru (2018) found that commercial facial recognition systems had error rates of up to 35% for darker-skinned women, compared to less than 1% for lighter-skinned men, due to intersectional disparities in representation in training and benchmarking datasets.

Want to explore more around bias in generative AI? Check out the tutorial Generative AI and Bias.  

Ask yourself: 
Could the AI in your school misread you or someone else, based on identity, context, or culture? Do you revise the output, check the facts you received, or accept and copy-paste with little thought?  

5. Ethics as Strategy or Substance? 

Frank Pasquale (2015) and Seele & Schultz (2022) both caution that talk of AI “ethics” can become a smokescreen. Behind the glossy values statements, there may be very little actual human oversight. 

When the rules of powerful systems are hidden, when decisions are automated but unaccountable, it becomes harder to challenge injustice. Without transparency around how these technical systems work, how they are trained and what safeguards exist, and without empirical information on their environmental and societal impacts, discussions of ethics are little more than branding. 

Links

• Artificial Intelligence and Competition │ Competition Bureau Canada (2023) 
• Atlas of AI │ Crawford (2021) 
• From Greenwashing to Machinewashing │ Seele & Schultz (2022) 
• Multistakeholder Perspectives on Military AI │ Sisson (2023) 
• Race After Technology: Abolitionist Tools for the New Jim Code │ Benjamin (2019) 
• Surveillance Capitalism and the Challenge of Collective Action │ Zuboff (2019) 
• The Black Box Society │ Pasquale (2015) 
• The Costs of Connection │ Couldry & Mejias (2019) 
• The Decolonial Turn in Data and Technology Research │ Couldry & Mejias (2021)
• Towards AI Ethics’ Institutionalization │ Schultz & Seele (2022) 

Discuss

Going Beyond Assumptions 

Complete this sentence for yourself, then share and discuss:

After what I’ve seen, I will no longer assume that AI in education is just… 

Use your responses as a springboard for discussion: 

• What new dimensions or risks came into focus for you? 

• Did the examples in the tutorial shift your view of AI tools you already use? 

• How might your role as a student, educator, or policymaker shape the way you think about AI’s purpose in education? 

Where should the boundaries be between educational use and other uses of AI infrastructure? 

What do you think? Tell us using the comment below.