The latest installment of the Rethinking Technology & Creativity column in TechTrends is published. This one, co-authored with Danah Henriksen, is called “The Mirror and the Black Box: AI Metaphors and What They Mean for Learning.” (Full citation and link below.)

The core question is deceptively simple: what metaphors do people reach for when they try to make sense of GenAI? We collected and organized dozens of them, from the purely mechanical (calculator for words, spicy autocomplete, blurry JPEG of the web) to the practically mythical (oracle, golem, sorcerer’s apprentice, Frankenstein). No other technology in history has generated a metaphorical spectrum this wide.

But the paper goes beyond taxonomy. What we argue is that something genuinely unprecedented is happening in the history of ideas. There is a pattern, traced in Draaisma’s book Metaphors of Memory, where every era reaches for its most impressive technology to explain the most mysterious thing it knows… the human mind. Plato used wax tablets. The hydraulic age gave us “humors.” Clocks gave us “gears turning.” Computers gave us “input,” “output,” and “bandwidth.” The direction was always the same: from the known and mechanical to the unknown and mental. Technology explained mind.

With GenAI, the arrow flips. For the first time, we use mind-language to explain a machine. We say it “thinks,” “learns,” “hallucinates,” “reasons.” And because the source domain (our own minds) feels familiar but is actually just as opaque as the target domain (the AI), we end up with what we call a double black box… one thing we don’t understand explaining another thing we don’t understand. The experiential familiarity of having a mind creates a feeling of comprehension that is entirely unearned.

The paper also mentions “honest non-signals” (building on Andrew Maynard’s work that I have written about here). I have been thinking about this idea a lot, and though I like the underlying idea, I am not sure that I like the term. I have discussed this with Andrew, and I am now learning towards calling this hollow signals. AI’s fluency is real, its helpfulness is real, its deference is real… but these properties don’t carry the same information they would coming from a human. Our epistemic vigilance, calibrated over millennia for a world where the only fluent, helpful agents were other people, has no protocol for this. It waves the signals through, without inspection.

The paper ends where it had to: with the argument that the most consequential metaphor isn’t about AI at all. It’s about learning. If we think of learning as information transfer, then AI looks like it does the job better. If we think of learning as cultivation, as apprenticeship, as the slow and difficult process of becoming… then AI looks like something else entirely. Same technology, radically different conclusions. The difference lies in the metaphor. Citation and link to the article below:

Mishra, P., & Henriksen, D. (2026). The mirror and the black box: AI metaphors and what they mean for learning. TechTrends. https://doi.org/10.1007/s11528-026-01197-y

Bonus: The Talmudic version

For those who want to experience the argument rather than just read it, I built an interactive website that presents the paper’s core ideas in a format inspired by the Talmud… the central argument as the core text, with commentary, connections, and reflections surrounding it. I wrote about the design process in an earlier post, “The Paragraph is the Interface: AI Metaphors meet the Talmud.” The website is at punyamishra.com/ai-metaphors. Check it out

Previous posts on metaphors

I have been thinking and writing about metaphors for a while now – not just in the context of AI. Here are links to some previous blog posts relevant to this topic

• The Mirror and the Machine: Navigating the Metaphors of Gen AI
• Shattered: Myth, Metaphor & Gen AI
• From Self-Driving Cars to Selfish Genes: Trapped in AI’s Metaphors, Literally
• Metaphors, Minds, Technology & Learning
• “Tipping” the Scales: When Metaphors (Quite Literally) Blur Reality
• Of metaphors & molecules: Bridging STEM & the arts
• Multiple metaphors & science learning: New article, new illustrations