10. The Stanford Marshmallow Experiment (Pt. 2)

On The Epistemic Status of Real versus Imagined Rewards

The published paper for the Stanford marshmallow experiment is called Cognitive and Attentional Mechanisms in Delay of Gratification. Contrary to popular conceptions, the focus of the study was only in part about the delay of gratification.

The main focus of the study is on attention and what differences it can have on our behavior. The paper outlines three experiments, each with multiple conditions. The results, generalized, can be stated as thus:

  1. The children with the marshmallow in plain view were only able to wait, on average, less than a minute to claim it.

  2. Those who were thinking about other things, even with the marshmallow in plain view, were able to wait longer than ten minutes on average.

  3. Those who were told to think about the marshmallow when it was not in plain view were only able to wait slightly longer (both less than a minute) than the children who had the marshmallow in front of them.

The third point is of most interest. There appears to be only a slight difference in capacity to delay gratification between thinking about the reward (a rather trivial marshmallow) and physically having the reward in front of the child. Both the imagined and the sensed marshmallow are “real enough” epistemically to incite behavioral changes to similar degrees. As far as our behavior is concerned, it seems it is the content of our attention that changes our behavior, not necessarily the ontological status or the presence of the reward. Moreover, it seems that our faculty of attention does not know, or care about, the difference.

If the precursor to the Stanford marshmallow test tells us that our behavior is moderated by attention, changing and modulating our behaviors; then this specific study tells us that our behavior can change for imaginary rewards nearly as much as visible rewards insofar as we focus our attention onto them. Coupled together, it follows that our behavior may be significantly changed and moderated by trivial rewards and this change in behavior occurs whether this trivial reward is held as a mental state — an imagining — or readily and physically available.

Now this is kind of a big deal. Our faculties of attention doesn’t care if the reward is real or not. And as long as we think about the promise of a future, trivial reward, it has a sufficient effect so that we cannot help ourselves from consuming it.

Knowing this, it should be no surprise why the apps on our phones, our computers, and the web sites we use are employing notifications to gain our attention. What we pay attention to becomes a “real enough” mental state of there being a marshmallow waiting for us. And so we act on them. Every notification we check is a new marshmallow test that we fail. We consent to notifications, falsely believing that they will save us time and effort. But what begins to happen is that we become addicted to the notifications themselves.

That is, the notifications become a marshmallow test of their own. When we check our phones, we play a meta-version of the marshmallow test where we give up and check our phones for a reward — the notification. When we see a notification, we play a sub-level of the marshmallow game where we check the notification to see if we have received some sort of trivial reward. For our faculties of attention, these rewards do not have to be real, and even when they are real, they can be utterly trivial. Still, there is something about them that when they catch our attention, we are unable to keep ourselves from consuming them for very long.


If I were a company effectively working as attention-brokers, I would find ways to develop ways to entrench this more embedded levels of a version of the marshmallow test. In fact, this is what they do.

From the Subcommittee on Antitrust, highlights mine:

By a former employee’s own testimony, a product manager’s only "job” is to hold your attention for 60 more seconds. Every single product team wants 60 more seconds of attention. Each minute can be directly mapped onto a monetization rate. Their entire task is to grab another minute of your time, because that simply means more revenue. It doesn't matter what they show you. It doesn’t matter how they do it. What matters is your additional 60 seconds of compliance. They have an entire arm of the company solely prepared to run behavioral experiments on you, to prod and to poke, to manipulate your emotions and thought patterns, just to get you to agree to staying on the platform for another 60 seconds. They keep and track these results. They know what works and what doesn't. And whatever works they work more and more to improve it. This is their product — to tear down your basic defenses, by hook or by crook, to continue to create marshmallow tests for you to fail. Because every time you fail one, means more revenue for these web sites.


If we accept the testimony is true, then it seems to reinforce the conclusion that the reason these organizations want our attention is because they change how what we think about and from there, change how we behave. Inversely then, it is logical that it is entirely possible to choose to no longer subject yourself to a regress of self-imposed marshmallow tests.

I will lay out my argument as such:

  1. Our thoughts, our mental content, and the thoughts we think cause us to act in ways we otherwise may not have acted.

  2. The thoughts we think are moderated by what we attend to — what we pay attention to.

  3. What we pay attention to moderates our thoughts, which moderate our behavior in the world. (1, 2)

  4. Assert that we are morally responsible for our actions if we could have chosen to do otherwise.

  5. We can choose to pay attention to something otherwise.

  6. ∴ We are morally responsible for our attention. (4, 5)

  7. If we choose to pay attention to something else, then we have some certainty in the notion that we could have done otherwise.

  8. If we can choose to pay attention to something else, then we have some degree of freedom to choose otherwise. (3, 5, 7)

  9. Our moral responsibility can be derived from our capacity to otherwise pay attention to something else. (6, 8)

  10. ∴ Therefore, being morally responsible for our attention is what leads us to becoming morally responsible agents. (4, 7, 8, 9)

  11. ∴ We now have license to conclude that we should be morally responsible for our own attention.


There is no question that each child in the Stanford experiments could have extended their wait time by another 2 seconds by paying attention to something else in the room. Then, it follows that if the child could have waited 2 more seconds, then the child could have waited another 5, 10, 15 seconds. The length of time that the child could wait is now up to the discretion of the child, had he chosen to pay attention to what he pays attention to.

Attention, being something we could choose to do otherwise in of itself, leads us to act towards different outcomes. Thus, we can choose to act otherwise if we choose to pay attention to different things. The hypothesis being that if things that divert our attention gets us to act in different ways, then surely the opposite is true.

If agency requires the capacity for attention, and attention moderates our behavior, then the way to reclaim agency comes from reclaiming our faculties of attention. And, I will later argue, as I have before, it is perhaps this capacity for attention control (not necessarily the amount exerted) that appears to be the most fundamentally human quality.