ResearchGate Twitter Flickr Tumblr

Learning “Trust” With Co-op Games

A circulating story about a grey parrot performing unexpectedly well in a setup of the classic 1972 Stanford Marshmallow Experiment reminded me of a related study from 2012, “Rational Snacking: Young Children’s Decision-Making on the Marshmallow Task Is Moderated by Beliefs About Environmental Reliability” (paywall). Takeaways from the latter were: the original test’s setup didn’t control for trust (and probably other factors as well), and the dependent variable of self-control is at least “moderated” by the perceived reliability of the environment—trust—in personal delay-of-gratification success or failure. (About the parrot, alas, I have nothing useful to say.)

One can be reasonably sure that both factors, i.e., trust and self-control, are indeed highly correlated in real-world settings. From there we can assume another confounding factor pertaining to both the original and the follow-up study that hasn’t been controlled for, namely, social background. Reliability isn’t just a matter of character, but also a matter of circumstances: people can be unreliable und untrustworthy not because they are unreliable and untrustworthy, but also because their personal environment and circumstances force them to be. For a taste of how this works, play the terrific Papers, Please!.

In fact, your unreliability and untrustworthiness in real-world settings can be a function of naked necessity, and the more precarious your personal circumstances are, the less you can afford being reliable and trustworthy. It can start with the promise of a birthday present you couldn’t keep because you needed to have the refrigerator repaired that just broke down, or for a visit to the zoo or your daughter’s all-important afternoon baseball game because you finally didn’t dare leave early from work even though you had asked for permission.

Now think again about the results of the original Marshmallow Experiment’s follow-up studies:

In follow-up studies, Mischel found unexpected correlations between the results of the marshmallow test and the success of the children many years later. The first follow-up study, in 1988, showed that “preschool children who delayed gratification longer in the self-imposed delay paradigm, were described more than 10 years later by their parents as adolescents who were significantly more competent.”

What needs to be done, actually, is factoring in the participants real-life environment reliability which, stunningly, hasn’t been controlled for throughout these studies.

Now, learning to trust or not to trust is a bit more complicated than “learning whom to trust or not” as an educational objective. “Trust” has a context, often a dynamic one. If children don’t understand this, they will have no clue what to do or what’s happening, e. g., when it comes to intermittent trustworthiness because circumstances compel a basically trustworthy person to be trustworthy most of the time, but not all of the time. And we know from Behaviorism that intermittent rewards (i.e, reinforcements) trigger certain classes of hormones that are not only highly addictive, but change behavior in unexpected and overwhelmingly socially negative ways.

When “Trust” as an educational objective has to take dynamically changing contexts into account, many of the game mechanics used in educational games are not even close to being useful, and that applies to corporate trainings as well. But how can we create dynamic contexts?

Well, Storification! A dramatic structure, turning points, ever higher stakes, unforeseen predicaments, and a “story world” where every action has its consequences, though not necessarily the expected ones. Obviously, in an interactive, participative game where the player can expect to have agency it’s neither possible nor desirable to “script” all these elements beforehand. As of now, the single best way to experience such situations, and experience them in perfect safety!, are still old-fashioned pen-and-paper roleplaying games—true, authentic “co-op” games through-and-through.

And that’s exactly what we’re going to have to create in the field of game-based learning: a plot-driven, context-rich co-op game that dynamically evolves through interactions between players who enjoy true agency. Which would be almost too easy with Strong AI as a “game master,” but as long as we can’t have that, we must work with what we have. Which, actually, is quite a lot—with MMORPG mechanics and mixed AI/Tutor systems leading the way.