01Literacy is an updating concept
The word "literacy" originally meant "the ability to read and write". Until the late 19th century, literacy rates in much of the world stayed below 20%. For someone who couldn't read, the contract, the newspaper, the Bible were all things that had to be "interpreted by somebody else". Being literate was, in itself, a place in the upper layer of the power structure.
As time progressed, the content of "literacy" was updated. Numerical literacy, media literacy, computer literacy, data literacy — and now, AI literacy. Each of these, at the moment of its emergence, drew a line between "those who have it" and "those who don't", and eventually became absorbed into society as a basic skill.
AI literacy is still at the stage where "those who have it" and "those who don't" are walking side by side. That is exactly the moment when dissecting its structure is most useful.
02A working definition of AI literacy
There is no single official definition of AI literacy yet. OECD, UNESCO, Stanford HAI, and various industry bodies have each proposed their own. Cutting across them, a working definition emerges:
AI literacy is the integrated set of cognition, skill, and attitude needed to use AI, understand it, judge it, treat it ethically, and create with it alongside others.
It isn't just "being able to use it". It isn't just "knowing how it works". The "eye to spot hallucination", the "judgment to decide when not to use it", the "ability to co-create with others through it" — only when all of these are present does it deserve the name "literacy". That is roughly where the field stands today.
03The structure ── five pillars
Decomposed concretely, AI literacy rests on the following five components:
Operational ability
Operating AI tools, writing appropriate prompts, and extracting outputs aligned with a goal. The most visible layer.
Comprehension
Knowing — at least at the conceptual level — how AI works. That it is a probabilistic model, that it depends on training data, that it has a context-window limit.
Discernment
Assessing the correctness and appropriateness of output. Hallucination detection, source checking, bias awareness, calibration of trust.
Ethical stance
Considering privacy, copyright, accountability, unfair use, and social impact. The sense that distinguishes "can do" from "should do".
Co-creation
Treating AI not as a "single answer generator" but as a dialogue partner with whom you iteratively build, alongside yourself and others. The most expertise-demanding layer.
Metacognition
The ability to notice, among the five above, where you yourself are currently weak. The self-observation that sits at the core of AI literacy.
04Classified as a competency
Sorting the five components into the KSAV model (Knowledge / Skill / Attitude / Value) used in education and HR makes clear why AI literacy is so complex:
| Attribute | Meaning | Primary mapping to the five components |
|---|---|---|
| Knowledge | Knowing how AI works, its limits, the regulations | Understand |
| Skill | Using, judging, combining | Use, Judgment, Co-create |
| Attitude | Constant verification, cautious handling, open-mindedness | Judgment, Ethics |
| Value | Fairness, accountability, patient/other-centeredness | Ethics |
| Meta-cognition | Reflective observation of one's own AI use and understanding | Transversal |
What this organization reveals: AI literacy cannot be completed by knowledge alone or by skill alone. A good programmer is not necessarily someone who can use AI ethically. A researcher who knows AI well may be poor at using it in the field. Only when KSAV plus metacognition mesh together do you get a literacy that means something at the societal level.
05What life looks like for those who "have" AI literacy
What will happen in the next ten years for people who have AI literacy? Concretely:
- Productivity at work: completing the same task in 1/3 to 1/5 the time. The gap accumulates.
- Learning speed: reaching a practical level in a new field in weeks rather than years, through dialogue with AI
- Quality of decisions: comparing multiple viewpoints instantly, reducing blind spots
- Filtering information: better odds of seeing the real thing among the noise
- Human relationships: delegating to machines what machines can do, freeing time for what only humans can do (empathy, generative dialogue, physical presence)
What matters is that these benefits do not come from "using AI" alone. Those who just use it get whipped around by it. The people who get results are the ones who have Judgment, Ethics, and Co-creation alongside Use.
06What happens to people who don't have AI literacy
Uncomfortable as it is, this part has to be written. What happens to people who lack AI literacy over the next decade:
Thinking itself gets outsourced to AI
When AI is used without Judgment, the output is taken at face value. Over ten years, the muscle of "thinking for yourself" atrophies. This is not a personal problem — it is a structural risk that the society as a whole loses thinking capacity.
The gap between "those who can ask AI" and those who can't exceeds the income gap
The 20th-century economic gap was about capital. The early 21st century was about digital access. From here on, it is about AI utilization capacity. Even with access to the same source, what you can extract differs by orders of magnitude.
A gap opens between "people who work with AI" and "people who get replaced by AI"
The middle tier — routine clerical work, junior analysis, formulaic content creation — is the most replaceable by AI. With literacy, you join the "users of AI" side. Without it, the "replaced by AI" side.
Voters who cannot judge "what is real" multiply
In an era flooded with generative-AI images, video, and text, those who cannot tell them apart are easier to manipulate at the ballot box and in public opinion. AI literacy ultimately connects to the quality of democracy.
07Four projected social splits
Building on the above, here are four splits I see coming:
- Education redesigns itself ── The center of gravity shifts from "memorizing knowledge" to "asking the right questions in dialogue with AI". The "memorize-the-right-answer" education ends; the "see-through-the-right-answer" education begins. Within ten years, university curricula will be substantially rewritten.
- The professions get redefined ── Physicians, lawyers, accountants, engineers. All were "people who had accumulated knowledge". From here on they are "people who, knowledge plus AI, can reach correct judgments". Professionals without AI literacy will rapidly lose competitive position.
- A new class structure ── Economists project a three-tier reorganization: "people who use AI", "people who are used by AI", "people who are replaced by AI". Mobility between tiers will be decided by AI literacy acquisition.
- "Thinking time" becomes scarce ── Ironically, AI was supposed to free up time, but the flood of stimuli reduces our deep thinking time. People who intentionally turn off the AI and protect time to think with their own head will become, in a new sense, "the privileged".
08What this means in pharma
This literacy discussion has concrete implications for the pharmaceutical field:
- HCP AI literacy: When physicians, pharmacists, and nurses differ in their ability to use AI, that gap connects directly to the quality of patient care. Training programs are no longer optional
- Patient AI literacy: This is the era when patients look up their symptoms in ChatGPT. The gap between those misled by errors and those using it appropriately affects adherence and timing of consultation directly
- Pharma company internal: MRs, marketing, R&D, regulatory affairs, quality — every department is now an "AI user". Internal AI-literacy training is becoming a required item, the same level as new-hire orientation
- Regulatory relations: Regulators will also split on AI literacy. Reviews by AI-literate vs. AI-naive officers will diverge in judgment. Industry now has a responsibility to contribute to the AI literacy of the regulatory side too
AI literacy is the "literacy rate" of a new era. Just as literacy rates determined a country's development a hundred years ago, AI literacy rates will determine organizational and national development over the next ten to twenty years.
But there is one thing the history of reading-literacy teaches us. Literacy did not spread by itself. Schools, libraries, public reading rooms ── those structures made literacy a social thing. AI literacy will be the same. Someone has to teach it. Someone has to make spaces where it can be learned. That responsibility, I think, falls on those of us alive at this turning point.
This site is one small attempt at that.