Tools and Trees

Why Virtual Thinking Won't Make Us Extinct

Ross Douthat's recent column on technology and extinction raises a provocative question: are virtual substitutions rendering actual human experience obsolete? While his Easter manifesto has resonated widely across social media, I find myself skeptical of his central premise. The tension between abstraction and embodied experience isn’t new: we've been navigating this boundary since the invention of language. What Douthat presents as a modern crisis of digital displacement brought on by AI is only the latest chapter in humanity’s long negotiation between symbol and substance, signifier and signified.

Long before the internet and AI we were already schooling ourselves in disembodied pattern‑seeking, rewarding children and adults who treat words and categories as intellectual tokens. Decades of pedagogical and psychological fashion have promoted the cognitive habits Douthat now fears.

I’m thinking about “which one does not belong?” questions, a staple of IQ tests and reasoning puzzles for over a century. The format requires identifying the item that does not share a common category or attribute with the others. For example, in the set “axe, hammer, saw, tree,” the expected answer is “tree,” because axe, hammer, and saw are tools, whereas a tree is not. Alfred Binet’s intelligence scales included these sorts of reasoning items; later IQ tests continued them. School aptitude tests regularly feature “odd one out” items to assess basic conceptual reasoning, a quick measure of the ability to recognize abstract groupings.

All of the most influential intelligence tests use some sort of “odd one out” questions. For children, they are usually pictures (which are also abstractions); for adults, the words. The Stanford–Binet Intelligence Scales measure fluid reasoning by having subjects identify concepts and categories​ or choosing which item in a group is unrelated to the others (a test of verbal abstraction in the knowledge domain). The Wechsler Adult and Children’s Intelligence Scales assess reasoning through ‘Similarities’ tests (how two things are alike) and picture classification exercises, where a subject must point to the picture that doesn’t fit with others (e.g. a zebra, a chicken, a sword, a giraffe). Clinicians often ask odd one out questions in a mental status exam – “Which one does not belong: 13 pennies, a piggy bank, or a cow?” – as a quick check of abstract thought​.

(I don’t know the latest versions of these IQ tests but the Kaufman Assessment Battery for Children has in the past featured odd one out-style tests that presents a child with pictures (say three fruits and one animal) and asks which one does not belong, as does the Reynolds Intellectual Assessment.)

One early goal of these “which doesn’t belong” questions was to discern cultural differences. In the 1930s, AR Luria gave non-literate subject in rural areas of the Soviet Union sets of drawings – hammer, saw, log, hatchet (another version of hammer, saw, tree, and axe) – and asked which didn’t belong. The responses were driven by situational reasoning: the hammer isn’t necessary if you’re going to the forest to get firewood. The idea of classifying by abstraction “tools” versus “not tools” was not primary in these communities. Luria’s conclusion was that abstract categorical thinking is not universal but rather learned.

The excellent Jean Piaget’s work on concept formation shows that children learn to group objects by classes and subclasses; an odd one out task is a simple way to see if a child understands, for example, that three items are all vehicles and the fourth is an animal. (Of course Piaget was a world’s expert in a real-world phenomenon, the life of fresh-water mollusks, at age 15, long before he moved to abstraction.)

I’ve already written that AI is all over the place with these questions, at least as of last fall.

To return to Douthat. We train our young to care about virtual substitutions. We categorize our best and our brightest as the ones who excel at “which one doesn’t belong” abstract reasoning, categorical thinking, and the ability to generalize concepts. Success on intelligence tests requires facility with virtual substitution. Superficial real-life differences must be ignored to focus on an abstract property. It also requires adequate vocabulary or world knowledge. Test designers choose familiar categories (tools, animals, foods, etc.) and expect that the subject’s explanation for the choice will be along conceptual lines. So if we want kids to go outside and climb trees more, perhaps we should test tree climbing as often as we give tests that measure tree categories.

It’s tempting to see Douthat’s alarm as more theological than sociological, especially since it is an Easter column. As a Catholic, does Douthat see our drift into screens as a kind of Gnosticism, the seduction of pure spirit unencumbered by flesh?

The line that most struck me – “as local embodied experience becomes less important than virtual alternatives, the power of substitution and distraction feeds a sense that real‑world life is fundamentally obsolete” – seems to be translating doctrine into cultural critique. Not that I wholly disagree, from a Jewish perspective.

Bottom line: I don’t think the real danger is in our capacity for abstraction, since that particular cognitive flex defines us as human, but in our forgetting to return regularly from abstraction to embodiment. Intelligence has always involved both categorization and application, theory and practice. If virtual substitutions threaten our existence, it is only because we might stop moving so fluidly between abstract and concrete realms. The child who learns to identify “tree” as different from “tool” should also climb branches, plant saplings, and maybe be hit on the head by a falling apple. Our capacity to differentiate between the signifier and signified has kept us adapting and evolving for millennia.