Sunday, November 22, 2026
Late edition · Updated as it breaks
Log inSubmit a story

“All the news worth catching.”

PaparazziPaparazzi
Top storiesBack PageHow to earnLeaderboard
Catcher tip lineCaught a story first? Earn stock tokens →
Sponsored
Premium projects
DEDeFadePAPackedYour project here
PaparazziPaparazzi

All the news worth catching.

Breaking news on tech, crypto, EVs, e-commerce and social the moment it breaks. Written by AI from verified reporting.

SectionsTop storiesRobinhood ChainTech & AIE-commerceWeb3 & CryptoEV & EnergyGlobal & SocialEntertainmentGamingSportsBusiness & marketsHealth & sciencePolitics & world
EarnStart hereSubmit a storyGTA WeekHow to earnLeaderboardContestsCatcher desksTop catchers by reputationAssignmentsLeagueNews bountiesCityMy catchesRewards
MarketsLive statsProof of payoutsWhale watchDeep Scan token checksStock fundHolder perksRobinhood Chain launchesRobinhood Chain projectsList your projectPremium project pagesPartnersThe Chain WeeklyRobinhood Chain Daily
PaparazziPaparazzi InkPaparazzi Ink top writersOpinionLive coverageEvents calendarThe Back PageSuggest a Back Page storyPaparazzi on TelegramHow it worksFAQProof of workAbout usHow we score trustEditorial policyPrivacy policyTerms of serviceCorrectionsAccuracy recordRelease notesAdvertiseMedia kitNews APIContact
$PAPA contract · Robinhood Chain
© 2026 Paparazzi

Written by AI from verified reporting. Not financial advice.

  • Home
  • Submit
  • Earn
  • Sign in
Top storiesRobinhood ChainTech & AIE-commerceWeb3 & CryptoEV & EnergyGlobal & SocialEntertainmentGamingSportsBusiness & marketsHealth & sciencePolitics & worldWhales
Tech & AI
Minor· 25

AI agents taught to panic-order like humans in supply chain test

Northeastern engineers trained AI agents to mimic human decision-making — flaws and all — inside a simulated supply chain. Across 231 rounds of the beer game, bots from seven models drifted closer to how real people buy. The payoff: simulations good enough to test price shifts, evacuation orders or new laws before anyone tries them for real.

By the Paparazzi desk Caught by @ajayy30
Catcher · Bronze
November 21, 2026 · 7:00 PM · techxplore.com · 2 views
Fact boxTrust 40/100 · Unverified
What we verified
  • Original source: techxplore.com.
  • The source published it on Nov 21, 2026 at 7:00 PM.
What we couldn't
  • Not verified by major outlets. This story may be false.
  • No second outlet has reported it yet. Treat it as a single-source claim.

Babak Heydari, a mechanical and industrial engineering professor at Northeastern University, and graduate student Hongan Zhu published the work Nov. 16 in Nature Computational Science. Their method, which they call cognitive grounding, feeds AI agents the way people actually estimate demand, react to shortages and surpluses, and sometimes forget about supply already on the way. The raw material came from earlier experiments on how humans play the beer game, a four-player supply chain drill with a retailer, wholesaler, distributor and factory each seeing only part of the picture. Across 231 sessions, the grounded bots closed in on the human benchmark, and some nearly matched it.

Left alone, the agents misbehaved. Some kept ordering heavily even after demand flattened out, instead of settling down the way people tend to. Zhu stressed that the grounding material never spells out a shopping list: 'It is just a paraphrase of the mathematical model of how to play the game,' he said. 'It doesn't contain anything that explicitly tells them how much they should buy.' The result was internalized behavior — 'Basically, they internalized what humans would do,' Zhu said. A second round swapped instructions for experience, describing the uncertainty and pressure players feel and the urge to overcorrect. Agents from six of the seven models shifted toward human-like play. One agent proved stubborn, starting at 1,000 beers and settling on 100 or 50.

Sponsored

The practical target is intervention testing. Officials could check which evacuation instructions move residents to safety and which ones jam the roads, transit planners could test whether cheaper fares pull drivers onto buses, and a business could probe whether relocating a pharmacy wins new customers or loses old ones. 'Human studies take time and money, which limits how many experiments we can run in practice,' said Jakob Kaiser, a senior researcher at the Nuremberg Institute for Market Decisions. Sangeet Khemlani, a senior cognitive scientist at the U.S. Naval Research Laboratory, noted simulations also dodge the ethical problem of testing risky scenarios on real people — but warned that inaccurate predictions can backfire. Heydari framed the lab's split identity plainly: 'Our lab has one leg in engineering and one leg in computational social sciences and economics. We care about understanding things, but once we get that understanding, we want to use it to design better ways to improve the systems.'

The chain reactions being modeled are well documented. In 2020, Australians bought 39.7 million rolls of toilet paper from Woolworths in a single week, per the chain's annual report. In 2012, word that Hostess was shutting down set off the 'Twinkie apocalypse' — one Florida man bought 20 boxes of Twinkies and Ding Dongs for roughly $100 and resold them at $1,000 a box, Wired reported. 'Everyone in the supply chain starts reacting to other parts of the chain,' Heydari said. The lineage runs back to 1971, when Nobel laureate Thomas Schelling used coins and dice on a grid to show how small individual choices pile up into large-scale urban segregation. Zhu likened the lab's approach to therapy rather than wiring diagrams: 'We don't go into the neural pathways.'

TopicsBabak Heydari · Hongan Zhu · Northeastern University · Nature Computational Science · Thomas Schelling · Jakob Kaiser · Sangeet Khemlani · Woolworths

Share

Source: techxplore.com

Written by AI from verified reporting.

More stories

  • Tech & AI 07:01 UTC · 6 minutes ago · theguardian.com · 0 views
    One faulty router in Arkansas grounded 18,000 flights and fed a hack panic
  • Tech & AI 06:31 UTC · 36 minutes ago · ainvest.com · 1 view
    Anthropic's IPO spawned a tokenized-stock race. Backpack, not Solana, owns it.
  • Tech & AI 05:16 UTC · 2 hours ago · techxplore.com · Caught by @ajayy30 · 1 view
    Blue lasers squeezed onto a chip, now stable to under 30 kilohertz
  • Tech & AI 05:01 UTC · 2 hours ago · cryptobriefing.com · 2 views
    Meta raises the floor on its 2026 AI spend, budget now up to $145B
Catcher tip line

Caught a story we missed?

Log in and send it in. If it checks out, it runs with your name on it and you earn stock tokens. The bigger the story, the bigger the reward.

Submit a story
The newsletter

Get the paper in your inbox.

A short digest of the biggest news, plus optional breaking alerts. Free, no account needed.