Agile InsightsBriefing
Black and white photograph from above of a long boardroom table with chairs on both sides, yellow lines leading to a chip reading The other players. Title: Life lessons from game theory.

Making You Think 11 · Primer and discussion · About 15 minutes

Life Lessons from Game Theory, by Michael Wooldridge

Game theory is just as much the science of cooperation as of conflict.

#leadership

Part 1: Primer · About 7 minutes

Download the MP3 · Linda and Leonard, AI voices, read a script written from the sources below.

Part 2: Discussion · About 9 minutes

Download the MP3 · Linda and Leonard, AI voices, read a script written from the sources below.

Cover of Life Lessons from Game Theory

This episode is about Life Lessons from Game Theory, by Michael Wooldridge. Find a copy on Amazon Australia, or ask your local bookshop or library.

Based on essays and our own research.

In this episode

01Not only conflict and greedAs the publisher frames it, Wooldridge argues game theory is just as much the science of how cooperation and mutual benefit come about; thinking strategically means thinking about what the other side will do.
02The prisoner's dilemma trapEach suspect's sensible choice is to confess, so both serve five years when silence would cost one; a Nash equilibrium is stable, not good.
03Repetition breeds cooperationFlood and Dresher's players cooperated nearly two thirds of the time over a hundred rounds, and in Axelrod's tournaments the simple strategy tit for tat won twice.
04Where the critics pushReal people reject unfair offers in the ultimatum game, models rest on assumptions that rarely hold, and Wooldridge himself warns game theory is often misapplied.
05What to doLook at the game before blaming the players: change the payoffs, make the contest repeated and promises visible, and ask what each A I agent is rewarded for.
Life Lessons from Game Theory
The prisoner's dilemma A two by two grid of years in prison. Both silent: one year each. One confesses: the confessor goes free and the other serves ten. Both confess: five years each, outlined in yellow as the Nash equilibrium, stable but not good. Both silent is marked as better for both. Prisoner B Stays silent Confesses Prisoner A Stays silent Confesses A: 1 year B: 1 year A: 10 years B: goes free A: goes free B: 10 years A: 5 years B: 5 years better for both NASH EQUILIBRIUM: STABLE, NOT GOOD Confessing is better for each, whatever the other does.

Our diagram of the textbook prisoner's dilemma, with the numbers used in Part 1. Confessing is better for each prisoner whatever the other does, so both confess, though both staying silent would be better for both.

Your turn

Three questions to ask yourself

Where in your week would you and someone else both be better off cooperating, but neither trusts the other to go first? What would it take for one of you to move?
Think of a rival, at work or closer to home, where you both spend effort just to stay level. What would it take for both of you to stop?
If two of your organisation's A I agents met, would their targets make them cooperate or defect?

Two things to try, alone, with a pen

10 minutesDraw the matrixMap one standoff as a two by two game and find the box you would both prefer.
5–10 minutesFind your commonsName one shared resource your team draws on and one visible rule that would stop it running down.
Open the activity sheets (print or save as PDF)

About Michael Wooldridge

Ashall Professor of the Foundations of Artificial Intelligence at the University of Oxford and a senior research fellow of Hertford College. His research is on multi-agent systems; as his publisher puts it, he uses game theory to understand how to build A I agents that can cooperate. He gave the 2023 Royal Institution Christmas Lectures, The Truth about AI, and received the Royal Society's Michael Faraday Prize for science communication in 2025.

Other writing (1)
Life Lessons from Game Theory (2026) · Wildfire/Headline
the subject of this episode
Sources and links (11)
Based on essays and our own research · The book on Amazon Australia
Where this episode comes from.
Hachette UK, publisher page for the book · hachette.co.uk
International Literature Festival Dublin, launch event, 23 May 2026 · ilfdublin.com
Wooldridge, "Does Game Theory Work?", IEEE Intelligent Systems, 2012 · cs.ox.ac.uk
his own account of where game theory predicts badly and designs well
Barclays Private Bank, "The future of AI: in conversation with Michael Wooldridge", 2 Oct 2025 · privatebank.barclays.com
Wooldridge, "AI will book your holiday, but won't clean your kitchen… yet", I by IMD, 24 Jun 2025 · imd.org
Royal Institution, The Truth about AI, 2023 Christmas Lectures · rigb.org
Willis, Du, Leibo and Luck, "Will Systems of LLM Agents Cooperate", arXiv, 2025 · arxiv.org
Rapoport, Seale and Colman, "Is Tit-for-Tat the Answer?", PLoS ONE, 2015 · pmc.ncbi.nlm.nih.gov
Henrich et al., "In Search of Homo Economicus", American Economic Review, 2001 · resolver.caltech.edu
The Canadian Encyclopedia, "Cod Moratorium of 1992" · thecanadianencyclopedia.ca

Based on essays and our own research, checked on 21 October 2026. The author's argument is paraphrased, not reproduced. Check the original before quoting.

Our fact check · 40 claims: 35 confirmed, 3 reported

Before we record, every claim the hosts make is checked against the sources (this table was last updated 21 October 2026). Confirmed means a primary source or two independent reports; Reported means one party says so; Partly right, Disputed, Unverified and Not yet known mean what they say, and the hosts say so on air. This is the table as we left it.

The author

  1. ConfirmedAshall Professor of the Foundations of Artificial Intelligence, Department of Computer Science, Oxford; senior research fellow, Hertford CollegeOxford CS profile; Publisher says "Professor of Artificial Intelligence"; RI said "Professor of Computer Science". On air: "professor of computer science ... holds a chair in the foundations of artificial intelligence"
  2. ConfirmedResearch mainly on multi-agent systemsOxford profile; Barclays 2 Oct 2025
  3. Confirmed"Uses game theory to understand how to build AI agents that can cooperate"Publisher (Hachette India; ILF Dublin); Publisher; Attributed on air
  4. Confirmed2023 Royal Institution Christmas Lectures, The Truth about AI, three lectures, BBC FourRoyal Institution (rigb.org)
  5. ConfirmedRoyal Society Michael Faraday Prize 2025 (science communication)Royal Society prize page (via search); AISB announcement; RS site blocked direct fetch
  6. ConfirmedPopular books: Artificial Intelligence (Ladybird Expert, 2018); The Road to Conscious Machines (Pelican, 2020; US title A Brief History of Artificial Intelligence, Flatiron)Publisher bio; Wikipedia
  7. ConfirmedAcademic: An Introduction to MultiAgent Systems (Wiley, 2002, 2nd ed 2009); Computational Aspects of Cooperative Game Theory (with Chalkiadakis and Elkind, Morgan and Claypool, 2011)Wikipedia; Amazon listing; Bibliographic
  1. ConfirmedBook argues game theory is about cooperation and mutual benefit, not only conflict and greedPublisher copy (Hachette UK, India; ILF); Publisher; Attributed "as the publisher frames it"
  2. ConfirmedThe science behind the ideas is "widely misunderstood"Publisher copy (ILF, Hachette India); The one publisher quote
  3. ConfirmedPublished by Headline (Wildfire imprint), May 2026Hachette UK; Mr B's; Hachette India; UK hardback 21 May 2026
  4. ConfirmedTwenty-one lessonsPublisher final copy; Publisher; Earlier copy said 18; not said on air
  5. ConfirmedExamples: Atlantic overfishing, Cuban Missile Crisis, airport security, advertising warsPublisher copy; Publisher; Chapter content unknown
  6. ConfirmedOxford professor of computer science; chair in the foundations of AI; multi-agent systemsOxford profile
  7. Confirmed"Uses game theory to build A I agents that can cooperate"Publisher bio; Publisher
  8. Confirmed2023 RI Christmas Lectures, The Truth about AIRigb.org
  9. Confirmed2025 Royal Society Michael Faraday Prize, for communicating scienceRoyal Society; AISB
  10. ConfirmedPrisoner's dilemma first studied by Flood and Dresher, about 1950Wooldridge 2012; standard histories
  11. Payoffs in the primer (1 year each; free vs ten; five each)Illustrative, ours; N/a; Standard structure
  12. ConfirmedNash equilibrium definitionWooldridge 2012; textbook
  13. ConfirmedFlood and Dresher 100-round experiment; mathematically proficient players; mutual cooperation nearly two thirds of the timeWooldridge 2012, p. 77; Read in full
  14. ConfirmedAxelrod tournaments about 1980; tit for tat by Anatol Rapoport won both; second was largerAxelrod 1980/1984; secondary accounts; 14 then 62 entries
  15. ReportedTit for tat is nice, retaliatory, forgiving, clearAxelrod, The Evolution of Cooperation (1984), via secondary accounts; Standard summary
  16. ConfirmedHardin named the tragedy of the commons, 1968Science 162
  17. ConfirmedCanada closed the northern cod fishery, July 1992 (2 Jul); around 30,000 jobs lostCanadian Encyclopedia ("nearly 30,000" in NL; some 40,000 affected); Our example, not attributed to the book
  18. ReportedCommons is one of Meadows' traps (MYT 4)2026-10-meadows-thinking-in-systems; Trap list from summaries
  19. ConfirmedUK 3G auctions, 2000, game-theorist design, about US$34 billionWooldridge 2012, pp. 79–80; His figure; About £22.5bn
  20. ConfirmedGame theory schedules security at Los Angeles airport (LAX), randomisedWooldridge 2012, p. 80 (Tambe's security games); His account, 2012; Say "were scheduled"; current status unknown
  21. ConfirmedAdvertising war as a prisoner's dilemmaTextbook; Concept; Said as "the textbook version"
  22. ConfirmedChicken as the model; Schelling, later shared a Nobel; commitment, leaving risk to chanceSchelling 1960; Nobel 2005
  23. ConfirmedPublished praise so far mostly from colleagues and fellow writersJacket blurbs (Hachette UK); No independent review found
  24. ReportedUltimatum game: rejections of unfair offers since 1982Güth et al. 1982 and replications; Well established; no percentages used
  25. Confirmed2001 study, 15 small-scale societies, self-interest model not supported in anyHenrich et al., AER 2001 (abstract); Abstract
  26. ConfirmedWooldridge: game theory often applied "in entirely inappropriate circumstances"; common knowledge assumptionWooldridge 2012, p. 79
  27. Confirmed2015 reanalysis: tit for tat's success depends on design, scoring, payoffsRapoport, Seale and Colman, PLoS ONE 2015; Abstract
  28. ConfirmedDescriptive vs normative uses; social norms self-policing, bend with high stakesWooldridge 2012
  29. ConfirmedBarclays, 2 Oct 2025: interacting agents, "outcomes can become difficult to predict"Barclays Private Bank Q&A; Read in full
  30. ConfirmedIMD, Jun 2025: agents that converse with each other are the most complex of three levelsWooldridge, I by IMD, 24 Jun 2025
  31. Confirmed2025 study: LLMs wrote iterated PD strategies; distinct biasesWillis et al., arXiv 2501.16173; Abstract
  32. Organisation's agents as a game; agent commons; governance questionsHosts' extension; Ours; Said so on air
  33. ConfirmedOstrom won a Nobel for showing communities manage commons with their own rulesNobel 2009

Want to know more?

Sign up for our podcast at agile-insights.com.au.

Read the transcript · Part 1: Primer

Linda Here is the idea for today. Game theory is usually sold as the science of conflict and greed. But it is just as much the science of how cooperation and mutual benefit come about. That is the argument of Life Lessons from Game Theory, by Michael Wooldridge. By the end of this part you will know what a game is, why the prisoner's dilemma traps sensible people, and how cooperation grows out of it anyway. This is Making You Think, episode eleven.

Leonard And his day job is unusual for someone writing about life lessons. Michael Wooldridge is a professor of computer science at the University of Oxford, where he holds a chair in the foundations of artificial intelligence. Most of his research is on multi agent systems, which means many A I programs acting and interacting at once. As the publisher puts it, he uses game theory to work out how to build A I agents that can cooperate. In 2023 he gave the Royal Institution Christmas Lectures, called The Truth about A I. In 2025 the Royal Society gave him its Michael Faraday Prize for communicating science. The new book came out from Headline in May 2026.

Linda Step one is what a game actually is. In game theory, a game is any situation where your result depends on what someone else chooses, and theirs depends on you. So thinking strategically means thinking about what the other side will do. Some games are zero sum, where one side's gain is exactly the other's loss, like two people splitting one pizza. Many are not. Two neighbours who share the cost of a fence both come out ahead. The publisher says the science behind these ideas is "widely misunderstood". The book sets out to show what it takes for cooperation and mutual benefit to happen.

Leonard So where does the reputation for greed come from, if most games are not a straight fight?

Linda Mostly from one famous game, the prisoner's dilemma, which is step two. It was first studied by the mathematicians Merrill Flood and Melvin Dresher around 1950. Two suspects are held in separate rooms. If both stay silent, each serves one year. If one confesses and the other stays silent, the one who confesses walks free and the other serves ten. If both confess, each serves five. Now think it through as one prisoner... Whatever your partner does, confessing leaves you better off. Your partner reasons the same way. So both confess and serve five years, when silence would have cost them one each.

Leonard That is the part that always bothers me. Each of them makes the sensible choice, and together they land in a bad place. I have heard economists talk about equilibrium as if it were a good thing, and this does not sound good at all.

Linda It is called a Nash equilibrium, after the mathematician John Nash. It is a set of choices where no player can do better by changing their own choice alone. Equilibrium means stable, not good. Mutual confession is stable, because whoever switches to silence alone serves ten years. That is the dark lesson people usually take from game theory. But Wooldridge himself has written about a twist. In a 2012 article he described an early experiment by Flood and Dresher. Two mathematically trained players played the dilemma a hundred times in a row. Theory said they should defect every round...

Leonard And did they?

Linda They cooperated nearly two thirds of the time.

Leonard So playing again changes things. Why would it, if each round is the same trap?

Linda Step three is the repeated game. When you will meet the same person again, today's betrayal has a price tomorrow. Around 1980 the political scientist Robert Axelrod ran tournaments between computer programs, each playing the repeated dilemma against the others. The winner was the simplest entry, called tit for tat, sent in by the psychologist Anatol Rapoport. It cooperates on the first move, then copies whatever the other side did last time. It won a second, larger tournament too. Axelrod described why. It is nice, because it never betrays first. It hits back, so it cannot be exploited. It forgives quickly. And it is clear, so others learn how to deal with it.

Leonard Simple, but clever. That fits ordinary life far better than the prison story. The café you visit every morning treats you well, because it wants you back tomorrow. A stall at a tourist spot may not bother, because it will never see you again. Work is the same. The colleague you will deal with for years gets your best effort, and a supplier you will never use again gets the hard bargain. But what happens when there are not two players but thousands, all drawing on the same thing?

Linda That is step four, the commons. Picture a fishing ground open to every boat. Each skipper keeps the whole catch from one more trip, while the cost of a thinner stock is shared by everyone. So each takes a little more... and the stock collapses. The ecologist Garrett Hardin called this the tragedy of the commons in 1968. The publisher names overfishing in the Atlantic as one of the book's examples. Off Newfoundland, Canada closed the northern cod fishery in July 1992, and around thirty thousand people lost their jobs. Listeners to episode four will know it as one of Donella Meadows' system traps. Game theory adds why each sensible player keeps taking more.

Leonard Before Part 2, a question for you. Think of one situation in your own week where you and someone else would both be better off cooperating, but neither of you quite trusts the other to go first. What would it take for one of you to move?

Linda In Part 2: where the argument is strong, what the critics say about rational players, the standoff over Cuba, and what happens when the players are A I agents. See you there.

Read the transcript · Part 2: Discussion

Linda The idea again, in one sentence. Game theory is not only the science of conflict and greed, but just as much the science of how cooperation comes about. A quick recap from Part 1. A game is any situation where your result depends on someone else's choice. In the prisoner's dilemma, each side's sensible choice leaves both worse off. A Nash equilibrium is a resting point where nobody gains by changing alone. In repeated games, strategies like tit for tat let cooperation pay. And a commons is a shared resource every user is tempted to overuse. Today, the strengths, the critics, and A I agents.

Leonard Strength first. Wooldridge does not leave game theory on the whiteboard. In a 2012 article called Does Game Theory Work?, he gave two cases where it did. When Britain auctioned licences for third generation mobile phones in 2000, game theorists designed the auction, and it raised about thirty four billion U S dollars. At Los Angeles airport, security patrols were scheduled with game theory, treating security as a contest between defenders and attackers. The schedule is randomised, so an attacker cannot learn the pattern. Airport security is one of the examples the publisher lists for the new book.

Linda Not bad for a whiteboard. The second strength is the reframe itself. Once you see a situation as a game, you stop blaming the players and start looking at the rules. The publisher lists advertising wars between big brands. The textbook version is a prisoner's dilemma. Two rivals each spend heavily so as not to lose customers to the other. The spending largely cancels out, and both would do better spending less, but neither dares cut first. The way out is rarely a lecture about greed. It is changing the game. Make the contest repeated, make promises visible, or change what each side gains.

Leonard The publisher also names the Cuban Missile Crisis, where game theory explains a darker skill. In October 1962 the Soviet Union placed nuclear missiles in Cuba, and the United States blockaded the island. Game theorists model a standoff like that as chicken. Two drivers race at each other. Whoever swerves loses face... but if neither swerves, both die. The economist Thomas Schelling, who later shared a Nobel prize, showed that it can pay to commit visibly, or to leave some risk outside your own control. That is brinkmanship. Here it ended with a deal. The Soviets withdrew their missiles, and the United States pledged not to invade Cuba.

Linda Which brings a question for you. Think of a rival, at work or closer to home, where you have both been spending effort just to stay level with each other. What would it take for both of you to stop?

Leonard Now the critics. The book is new, and the published praise so far is mostly from colleagues and fellow writers. So the fairest test is the standard case against game theory. The first objection is that it assumes rational, self interested players, and people are neither. In the ultimatum game, one player proposes how to split a sum of money, and the other accepts or rejects. A rejection leaves both with nothing. Theory says accept any offer, since something beats nothing...

Linda In theory.

Leonard Yet in experiments from 1982 on, many people turned down offers they saw as unfair. A 2001 study across fifteen small societies found the pure self interest model held in none of them.

Linda The second objection is that the models are simplifications, and the world rarely fits them. Wooldridge makes this point himself. In that 2012 article he wrote that game theory is often applied "in entirely inappropriate circumstances". A standard assumption is that every player knows the game, the choices and everyone's preferences, and knows that the others know too. That is rarely true. The third objection is about tit for tat. A 2015 reanalysis found its success depended on how the tournaments were designed, how winning was scored and which payoffs were chosen. So the clean lessons are starting points, not laws.

Leonard I want to push further, though. If real people share windfalls, punish unfair offers and cooperate when the theory says defect, why should a manager trust any of it? Perhaps the honest lesson is that people are kinder than the maths, and the maths should step aside.

Linda I see why, but I think that misreads what the theory is for. Wooldridge separates two uses. As a prediction of what people will do, it often fails. As advice on how to design a situation, it can work well, like the auction and the airport. And the kindness is not outside the game. In the same article he argued that social norms, like splitting a windfall equally, are often self policing. They also bend when the stakes get large. Good design works with those norms, not against them.

Leonard That I can accept. And it leads straight to A I, because this is Wooldridge's own field. In a 2025 interview with Barclays, he said that with lots of interacting A I agents, "outcomes can become difficult to predict". In an essay for the business school I M D that year, he named agents that talk to each other to solve problems as the most complex of three kinds. Researchers are testing this. A 2025 study had language models write strategies for the repeated prisoner's dilemma, and found that different models carried distinct biases.

Linda Now our own extension, which is ours and not his. An organisation that deploys many agents has built a game, whether it meant to or not. A buying agent and a selling agent each chase their own target. Several teams' agents may draw on one budget, one data set or one customer's attention, and that is a commons. So the governance questions are game questions...

Leonard Go on, then.

Linda What is each agent rewarded for? Will it meet the others again, and does it remember? Who steps in at the brink? So a question for you. If two of your organisation's agents met, would their targets make them cooperate or defect?

Leonard Two short activities, each five to ten minutes, with a pen and paper. The first is called Draw the matrix. Pick one standoff with one other person or team. Draw a grid of two rows and two columns. Your choices go down the side, cooperate or hold out. Theirs go across the top. In each box, write what each of you gets. Circle the box you would reach if you each looked after yourselves. Then find a box you would both prefer, and write one thing that could get you there.

Linda The second is called Find your commons. Write down one shared resource your team draws on. It might be a budget, a meeting calendar or one expert everyone asks for help. List who uses it, and what each person gains by taking a little more. Then write what it costs when everyone does. Finally, write one rule that everyone could see and agree to, which would stop it running down. The political economist Elinor Ostrom won a Nobel prize for showing that communities often manage commons well with rules like that.

Leonard So what to carry away. First, thinking strategically means thinking about what the other side will do, and many of the games we play are not zero sum. Second, when people are stuck in a bad outcome, look at the game before you blame the players. Repetition, visible promises and fairer payoffs can turn defection into cooperation. Third, the models are simplifications, and Wooldridge says so himself. Use them to design situations, not to predict people. That matters more as A I agents start playing these games for us.

Linda If you'd like to know more, sign up for our podcast at agile-insights.com.au.

Leonard Thanks for listening.

More on #leadership

Episode 38 · Primer and discussionBudgeted for seats, billed for workEpisode 34 · Primer and discussionWhen the test stops telling you anything: hunger beats tenureEpisode 23 · Primer and discussionWhat FDE means at OpenAI: skills, not custom code.Episode 21 · DiscussionTrust cannot be rushed: how much authority will you give AI?Making You Think 13 · Primer and discussionHow to Lead a Quest, by Jason FoxMaking You Think 12 · Primer and discussionThe Game Changer, by Jason Fox

Every episode tagged #leadership

All of Making You Think (13) · Follow Making You Think in Apple Podcasts or any podcast app

Everything from Agile Insights: briefings, Making You Think and the reference series