Skip to content
#142 The Games We Play 🎧 episode artwork

Sep 5, 2021 27 min

#142 The Games We Play 🎧

from Anticipating the Unintended Hosted by Pranay Kotasthane

Matsyanyaaya: Not So Great Game Theory in Afghanistan Big fish eating small fish = Foreign Policy in action — Guest Post by Ameya Naik US intelligence agencies considered it likely that the Taliban would retake control of all or most of Afghanistan following US withdrawal. Their estimate, however, was that this would take weeks or even months; the idea that Kabul would fall in a matter of days was considered a worst-case scenario. Now that the worst case has played out, analysts are scrambling to explain why. One narrative thus has it that ANA was so poorly-trained - and its leadership so corrupt - that once U.S. military contractors left the field, Afghan forces had neither the motivation nor the acumen to resist the Taliban advance. As President Biden himself put it "...we could not provide them... the will to fight." To test that claim, we can turn to Game Theory. Game Theory stylises any decision involving two or more players as a "game". * Each player can receive some "payoff" (or outcome) from the game. * Payoffs are "contingent", i.e. the payoff a player receives depends partly on their own decisions, partly on the decisions of the other player or players. * One assumes players will act rationally, making choices that maximise their payoffs. * Thus, if we know the payoffs each player can receive, we can predict their choices, and thereby also the outcome of the game. The most famous example of this is the Prisoner’s Dilemma: a game in which two people suspected of committing a crime are being questioned independently. If both deny committing any crime, the authorities will only be able to convict them of a minor offence (slightly bad outcome). However, if one of them bears witness against the other, the authority will let the snitch go free (best outcome) while convicting the other of a major offence (worst outcome). The textbook prediction is that both suspects will crack, and wind up with the worst combination of outcomes. However, the classical Prisoner's Dilemma is limited in one important way: it is a single-turn game. The players only make a choice once, at the same time, without knowledge of each others' choices - and then the game ends. A more realistic scenario is what is known as a repeated game. A repeated game has multiple turns: the same players interact, under the same rules, with knowledge of the choice made by the other players in the previous turn. Imagine that the suspects are schoolchildren and the authority is the Principal of the school. No one is going to jail; even if neither child knows what the other is saying to the Principal, they do know they will both interact on multiple occasions thereafter, both in class and outside it. Even with the same payoff structure, the outcome starts to look different: knowing that we have to meet the other person every day makes us far less likely to crack - or "defect", in game theory jargon - because they could punish our defection on the next turn. In a repeated game, players make decisions "in the shadow of the future". The political scientist Robert Axelrod modelled a repeated game in The Evolution of Cooperation. He demonstrates that the optimal strategy for such games is what he calls "Nice Tit-for-Tat". * Start by complying (being nice to the other player) * If they comply, continue complying (happily ever after) * If they defect, punish them by defecting on your next turn. * If they respond to your defection by complying, they have accepted the punishment, so go back to complying again. * If they defect again, you should also defect again (Tit-for-Tat) - a downward spiral until and unless they switch to complying. A key insight from Axelrod's work is that this strategy only works if the total number of turns is unknown. Why? * If the number of turns is known, we can try to pull a fast one - complying until the penultimate turn, but then defecting on the very last one (turn N), when there is no possibility of punishment thereafter. * The other player is not a fool. They know we are likely to defect on the final turn, so they will take precautions: they will defect preemptively, on the penultimate turn (N-1). * Since we know they will do this, we will defect on the turn before that (N-2). * They know we will do this, so they will defect on the turn before that one (N-3) - and so on till the whole chain unravels. The deterrence of future punishment only works if the number of turns is unknown. We can think of this as an infinite game or at least an indefinite one. What does all this have to do with Afghanistan? When President Trump committed to pulling out U.S. troops by a specific date, and then when President Biden made clear he would uphold that commitment (even with a different date), they converted an infinite game into a finite one. An open-ended U.S. presence was a signal of potentially unending US involvement, complete with punishment for behaviour the US considered unacceptable - for instance, overthrowing the US-supported government in Kabul. Afghan leaders broadly shared the assessment of the U.S. intelligence agencies: the Afghan government could not survive if US support was withdrawn. (Whether this is objectively true is beside the point; it seems to have been the mental model of the Afghan provincial governors, military leaders, and in all likelihood Afghan soldiers themselves.) With withdrawal confirmed, why bother resisting - especially given the Taliban's inhuman tactics, including targeting family members of soldiers, and threatening reprisal killings once they take power? Once the average Afghan believed that a Taliban victory was inevitable, the finite game unravelled. Players chose to defect (surrender / retreat / literally defect to the Taliban) at every step, and the timeline towards the fall of Kabul accelerated dramatically. Biden is precisely wrong. The US was providing Afghan leaders and forces with the will to fight, not by training and equipping them, but by making the prospect of a Taliban victory impossible. Given U.S. domestic sentiment favoured withdrawal, a better question might have been: even without a U.S. military presence on the ground, could the US establish deterrence against the Taliban? India Policy Watch #1: NMP, Another Gamechanger? Insights on burning policy issues in India — RSJ The Union government this week announced a National Monetisation Pipeline (NMP) with the aim to unlock value in existing infrastructure projects. The idea is simple and draws from the pioneering “Asset Recycling Initiative” done in Australia between 2013-16. Select assets that are already generating revenues (like roads, railway stations, power plants), lease them out to private sector bidders for a defined period, transfer the revenue rights to them, take an upfront payment for the lease, work out some revenue share arrangement on an ongoing basis, and have a few checks and balances to ensure the private sector doesn’t gouge the consumers on pricing or runs the asset to the ground over the years. That’s it. The government expects to raise ₹6 lakh crores in the next four years from this which it will use to fund greenfield infrastructure projects. To make sense of this number, the Union budget size this year was around ₹35 lakh crores. In the speech, the FM had reiterated the intent to spend ₹110 lakh crores over the next 4 years to create a National Infrastructure Pipeline. About 85% of that amount was to be raised through the traditional sources of capital (government borrowings) and the balance was to be taken up through innovative mechanisms. One such mechanism was the creation of a new Development Finance Institution (DFI) which would build a lending portfolio of ₹5 lakh crores in three years. The other mechanism is the NMP announced this week. So, what do I think of this? Let’s look at the reasons for doing this. * Our economic engine was slowing down even before the pandemic. Things have gotten worse since. The government can manage to keep its base in thrall with its favourite social and cultural… This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit publicpolicy.substack.com

Episode details come from the podcast's RSS feed.

Embed this episode

Ready to play

#142 The Games We Play 🎧

0:00 27:04

No transcript yet.

Requests help us decide what to transcribe next.

Frequently asked questions

How long is this episode of Anticipating the Unintended?

This episode is 27 minutes long.

When was this Anticipating the Unintended episode published?

This episode was published on September 5, 2021.

Can I download this Anticipating the Unintended episode?

Yes. Use the download control on the episode player to save the publisher-provided media file.