Department of Information Science and Applied Artificial Intelligence · Bar-Ilan University

DA'AT Forum

From Generative AI to Regenerative AI

Every week, the world measures what machines know. Almost no one measures what happens to what people know.

DA'AT is a forum of the Department of Information Science and Applied Artificial Intelligence at Bar-Ilan University. It focuses on a question that will shape the coming decade.

Join the monthly newsletter
The DA'AT logo: a tree whose crown takes the shape of a brain grows from an open book, with roots that turn into a network of digital data, above the words DA'AT, Regenerative AI.
The usual question

"Does the AI answer correctly?"

We ask

"What does AI leave in the hands of people and society after it answers?"

Why this matters now

Learning
+48%practice−17%test without AI

In a field experiment with about 1,000 high-school students, free access to GPT-4 raised practice scores by 48%. On a test without AI, the same students scored 17% lower than peers who never used it.

Bastani et al., PNAS, 2025
Science
−5%topics−22%interaction

Researchers who use AI tools publish more and are cited more. Yet the range of topics science explores shrinks by about 5%, and interaction between researchers drops by 22%.

Hao et al., Nature, 2026
Shared knowledge
−25%Stack Overflow−8%Wikipedia views

Within six months of ChatGPT's launch, activity on Stack Overflow fell by at least 25%. In 2025, Wikipedia reported a drop of about 8% in human page views.

Israel
38%Israel33%US17%China

38% of Israeli adults use AI tools, compared with 33% in the US and 17% in China. That makes Israel a natural laboratory for these questions.

Our vision

A society in which every advance in artificial intelligence becomes an advance in human knowledge.

DA'AT is building a new field: the sustainability of human knowledge. It treats human knowledge the way environmental science treats natural resources, as capital that can be measured, worn down and renewed. That capital exists at three levels:

Personal

What each of us knows, understands and can do on our own.

Collective

The diversity and independence of thinking across a society.

Shared

The sources, creators, languages and heritage that feed all of us, and the machines too.

Regenerative AI is AI that, on balance, adds to this capital. It leaves people more capable, keeps a diversity of views alive, and sustains the sources it learns from.

What the forum does

  • Develops a shared framework for studying knowledge sustainability, with research hypotheses that can be tested and proven wrong.
  • Launches joint pilot studies, drawing on Bar-Ilan's labs in behavioral and physiological research.
  • Rethinks how AI is evaluated: by what it leaves its users. That means independence, calibrated trust, diversity of views, faithfulness to sources, and control.
  • Hosts a monthly research seminar, working groups, workshops and visiting researchers.

Questions we explore

  • When does AI help build skill, and when does it create "knowledge debt"?
  • Does "the AI said so" replace the original source in our memory?
  • Do personal AI agents quietly shift their users' preferences?
  • When millions rely on similar models, do the public's errors start to line up?
  • Will signs that public knowledge resources are wearing down appear in Hebrew before English?
  • Can we train models to optimize for measured human growth rather than momentary satisfaction?
  • Can an interface inspired by the layout of the Talmud page, which keeps sources and disagreement visible, strengthen independent thinking?

What you'll get: the DA'AT monthly newsletter

  • Research digest: carefully selected findings on how AI is changing learning, memory, judgment, trust and diversity of thought.
  • First access to invitations for our monthly seminar, workshops and forum events.
  • Inside the research: updates from Bar-Ilan researchers and guest scholars.
  • Practical takeaways for educators, managers, AI developers and policymakers on adopting AI that builds expertise instead of eroding it.
  • Ways to get involved: working groups, collaborations and taking part in studies.

One email a month. No spam. Unsubscribe anytime.

The forum is coordinated by Prof. Maayan Geffet-Tamir, Department of Information Science and Applied Artificial Intelligence, Bar-Ilan University.

Join the newsletter

One email a month. No spam. Unsubscribe anytime.