Gabay et al. 2019 – MDMA and cooperation

Sun 19 July 2020

Here's a big question: can intentional psychedelic use create enduring behavior change here in the everyday world?

For helping with mental health issues like addiction, depression, and PTSD, the provisional answer seems to be "Yes, yes it can."

But what about for simpler stuff? Can psychedelics help us be kinder & more cooperative in our everyday lives?

Enter Gabay et al. 2019, an RCT where participants took MDMA, then played iterated prisoner's dilemmas while in a brain scanner. (Science!)

The prisoner's dilemma is a canonical piece of game theory. Here's a quick recap:

• You find yourself in a situation (let's say... being questioned at your local police precinct) where you can choose whether or not to cooperate with your partner in crime, but you can't communicate with them as you decide (they're being questioned in a different interrogation room)
• If you choose to cooperate (you stay silent) but they don't cooperate back (they rat you out), you receive an intense punishment (a 30-year sentence! You two have been up to some no-good stuff)
• If you don't cooperate (you rat them out), but they cooperate with you (by staying silent), they receive the intense punishment (they get the 30-year sentence) and you get to go free
• If you both cooperate (stay silent), you both get a reduced punishment (10 years in prison apiece)
• If neither of you cooperate (you both rat each other out), you each get a heavy punishment, though not quite as intense as you would have if you were taking the fall all on your own (20 years apiece)
  
  

Here's a chart representing the situation:

So what's the best thing to do? That's the dilemma.

Ideally you would both cooperate – that yields the best outcome for both you & your partner. It's the smallest amount of punishment overall.

But if you just consider your own situation, it might better to throw your partner under the bus (if you think they're going to cooperate) – if they stay silent, you would get to go free! However, if your partner follows that same logic, you would both receive a 20-year sentence.

This is a toy example, but the dynamic shows up everywhere. We're constantly confronted with situations where it'd be better overall to act in one way (telling the truth, showing up to meetings on time, following through on commitments), but better for us in the moment to act in a different way (a white lie to spare someone's feelings, being late, breaking commitments when convenient).

The stakes usually aren't as high as a 30-year prison sentence, so you get to play the game repeatedly, often with the same set of people. Over time, you get a feel for the different people you're playing with – some seem to be mostly cooperative, others only looking out for themselves. In game theory, this is modeled as an iterated prisoner's dilemma.

It's not a big stretch to imagine us all enmeshed in a web of interlaced iterated prisoner's dilemmas – a continual stream of decisions about whether we're going to cooperate with others in good faith, or instead find a way to defect & save ourselves. The aggregation of all these decisions is the world we end up living in.

We want to end up in the world where we're mostly cooperating with each other – that's better for everyone, by definition. Unfortunately, the rationally self-interested thing to do in an iterated dilemma is to defect, though the best move superrationally is to cooperate. (Superrationality is a bit of a slippery concept... it essentially means imagining yourself as one of a group of rational decision-makers, each of whom is imagining itself as one of a group of rational decision-makers.)

Because we want to end up in cooperate-world, an important project is finding ways to move out of an equilibrium that assumes game-theoretic rationality and to move towards an equilibrium that assumes (and cultivates) superrationality. If we could develop methods of reliably cultivating superrationality (or heuristics that approximate it), that would be extremely helpful.

Might MDMA be such a method? That's what Gabay et al. was investigating. Participants played six 15-round prisoner's dilemmas with three different kinds of partners – a trustworthy human who mostly cooperated, an untrustworthy human who mostly defected, and a game server who made a random decision each time. (In reality the "human" partners were also computer programs, but the participants apparently believed they were playing against other humans.)

Importantly, the trustworthy & untrustworthy partners weren't labelled as such... participants just saw that they were paired with "Greg" or whoever, and then had to discern the trustworthiness of their partner as they played.

Here's what they found:

The green dots (to the left) are the placebo results, the red dots are the MDMA results.

A few things stand out:

• There's not much difference between the placebo & MDMA sessions when partnered with a game server, and the amount of cooperation for the game server is lower than when partnered with a "human." This makes sense! If you're paired with a random-choice generator, you have no way to signal that you want to cooperate – it's just going to make a random choice each round. You might as well try to get what you can for yourself, seeing as there's no way to build trust.

• Both the MDMA & placebo groups cooperated more when paired with a "human" partner, even an untrustworthy one. When playing with a human, you can signal your intentions to cooperate.

• There wasn't a significant difference between the MDMA & placebo groups when playing with an untrustworthy partner, but there was a significant difference when playing with a trustworthy one – in that situation, the MDMA group cooperated more.
  
  

That last point is so interesting.

I think there's a worry floating around that drugs, especially drugs like MDMA, will make you uncritical, dull, and dumb. You might grow more content & peaceful, sure, but at what cost?

That doesn't seem to be the case here. Participants on MDMA were more likely to cooperate when they thought they were paired with a cooperative player, but no more likely to cooperate when they thought their partner was defecting. They were still able to discern which players were cooperative while on MDMA, and they were able to protect themselves from being exploited when working with untrustworthy partners.

Gabay et al. isn't the final word on MDMA and cooperation – it's a small study that has limitations, and other papers have found contrary results (we'll talk more soon about how we might roll all these together into an overall picture about what's going on here).

But it's such a striking result – perhaps MDMA can boost goodwill & cooperation without a concomitant increase in naïvety. That would be a powerful thing.