1. Situated Knowledge → Embodied → Relational → Transparent (Context Matters in AI Use)
* AI outputs are not universal truths; they’re situated in data, prompts, cultural and relational neural net probabilities that produced them. Knowing the context opens space for human/ai co-presence and co-design with AI for knowledge we can call more trustworthy.
* Example: A generative AI “explainer” of a historical event reflects training data and bias; literacy means asking whose perspective is being output.
* AI literacy skill: Question context:, relationships and transparency of training data, cultural frame, intended use-case.
2. User-Centered Ethnography → Observe Real-World AI Interaction
* Like her Xerox PARC copier study, AI literacy means studying how people (students, workers) actually use AI tools, not just how designers think they should.
* Example: In a university setting, watch how students actually integrate ChatGPT into research—copy-paste vs. iterative query refinement.
* AI literacy skill: Learn by observing and documenting real AI workflows to reveal gaps and affordances between design and use.
3. Sociomaterial Assemblages → AI is Human + Machine + Context
* AI isn’t a standalone intelligence; it’s embedded in sociotechnological systems—people, politics, infrastructures.
* Example: Facial recognition AI isn’t just “software”—it’s policies, cameras, training data, and enforcement practices.
* AI literacy skill: Map the whole system—data, code, humans, governance, expectations—before evaluating AI’s “intelligence” or fairness.
4. Figuring the Human → AI Mirrors Our Assumptions
* AI models often double human traits, amplifying both insight and prejudice.
* Example: When AI generates “professional” headshots, who looks most ‘professional’ in its output? What cultural norms are normalized, what others effaced? Where does your real face and mask fit in this picture and why?
* AI literacy skill: Treat AI outputs as a funhouse mirror—ask “What does this reveal about me?” How does trusting this reflect how I should trust societal expectations .
5. Critique of Autonomy → Demystify ‘Fully Autonomous’ AI
* No AI is fully independent; human design, training, and oversight are always in play.
* Example: “Autonomous agents rely on human categories, model limitations, and system configuration (thinking time, data searched).
* AI literacy skill: Replace “autonomous AI” with “thinking human-in-the-loop system” in your mental model.
6. Imaginaries of Omniscience → Question AI’s ‘All-Seeing’ Claims
* The myth that AI can achieve real-time total situational awareness now fuels hype and harm.
* Example: ask an actual programmerr how many mistakes an AI made in helping with his or her project. How does their comments jive with industry benchmarks
* AI literacy skill: Scrutinize AI claims—look for blind spots and uncertainty.
7. Thingness of AI → Keep AI as a Moving Target
* Once AI is seen as a general “thing,” it’s harder to critique. Keep it open, plural, and contestable.
* Example: Don’t just say “The AI says…”; specify which model, trained on what, and how the prompt was phrased.
* AI literacy skill: Always name the system, version, and provenance—resist treating AI as a singular, mystical 'all seeing' object.
Integration for 2025
Suchman’s message reframed for AI literacy is this: Work with AI as a situated, embodied, material practice. Every interaction with AI in 2025 is an opportunity to:
* Situate Knowledge as Embodied → Relational → Transparent. Understand the deeper context.
* Observe real-world patterns.
* Map the human + machine assemblage.
* Keep the AI provisional, and accountable.