The Marshmallow Trap: Why Our Bias for Action is Burning Us Out

You spoke up in the meeting? Marshmallow. You posted a long Slack update with five bullet points and a hasty plan? Marshmallow.

There’s an odd bias in many workplaces (especially in startups, and fast-scaling tech-adjacent businesses) where research is considered adjacent to work rather than work itself. You can spend a day reading white papers, auditing competitors, or digging through analytics dashboards, and someone will inevitably ask, “So, what did you do today?” If the answer isn’t a deck, a design, a prototype, or similar you may here, “So no work.”

Some people love to see manic flurries of activity regardless of the productivity resulting from such actions. A manager with this craving for action is an uphill battle. Initiatives are judged by both their success and how quickly actions are taken. In such environments, a lot of effort is wasted in the name of expediency.

The irony is that bad research (or no research) is one of the most expensive things a company can incur. When people skip the foundational legwork in favor of quick execution, they often end up executing the wrong thing. This results in burnouts, pivots, and retrospectives filled with words like "misaligned" or "premature." But despite this cycle repeating itself like a sitcom rerun, the idea that research isn’t real work remains firmly entrenched in some organizations.

Slow is Smooth, Smooth is Fast

The Department of Government Efficiency (DOGE) applied to government what I consider to be the most bass ackwards notions of startup land, including the bias for activity over results. Around a fifth of the jobs slashed at the Department of Health and Human Services were a mistake and will need to be corrected said Robert F. Kennedy Jr. United States secretary of Health and Human Services.

“Personnel that should not have been cut, were cut,” Kennedy told reporters on Thursday. “We’re reinstating them. And that was always the plan. Part of the DOGE, we talked about this from the beginning, is we’re going to do 80% cuts, but 20% of those are going to have to be reinstated, because we’ll make mistakes.”

Milman, O. (2025, April 4). RFK Jr says 20% of DOGE’s health agency job cuts were mistakes. The Guardian.

When you do work, do you enjoy knowing that you had an error rate of 1 out of 5 and that you will need to go back and correct these mistakes?

Errors are unavoidable. That said, an error rate of 20% for nearly anything is borderline sabotage. What is more, firing and rehiring people take time, and incur costs across a host of variables. High stakes situations call for caution.

The DOGE approach seems to glorify recklessness as long as something is happening. The plan, if we’re calling it that, was to make sweeping decisions quickly, accept that a fifth of them would be wrong, and then circle back to fix the mess later—under the pretense that this is some novel form of agility. But moving fast and breaking things only works even in theory when the things being broken are websites, not people’s lives, livelihoods, or the critical infrastructure of public health.

In most industries, if you knowingly implemented a strategy with a 20% failure rate and no robust process for preemptive accuracy, you'd be fired. In DOGE-land, it’s just Tuesday. The real irony is that a few weeks of research, audits, and scenario modeling could have dramatically lowered the error rate. But research doesn’t count—until the cleanup crew arrives.

Stanford marshmallow experiment

This allergy to patience isn’t just a quirk of modern management, it’s a measurable cognitive bias. Take the famous Stanford marshmallow experiment. In the 1970s, psychologist Walter Mischel offered children a deal: one marshmallow now, or two if they waited 15 minutes. The kids who were able to delay gratification, who sat, squirming in their chairs, not eating the marshmallow, were later shown to have better life outcomes: higher SAT scores, lower rates of substance abuse, and even better stress tolerance.

Restraint matters. Being able to pause and think before acting, especially in complex or high-stakes environments, pays dividends over time. But in many orgs, acting immediately gets you the marshmallow. You spoke up in the meeting? Marshmallow. You posted a long Slack update with five bullet points and a hasty plan? Marshmallow. You waited a week to gather data before giving a recommendation? That’s suspiciously marshmallow-free behavior.

What gets overlooked is that the second marshmallow, more accurate decisions, smoother rollouts, fewer emergency meetings, is always the result of discipline. Research is the pause. It's the squirmy 15 minutes of sitting with uncertainty, asking better questions, and resisting the urge to just do something for the dopamine. We tell ourselves we want strategic thinkers, but we reward reactive ones. So we get leaders who move fast and break things, and somehow never stick around to pick up the pieces.

And in that way, we’ve built a professional culture that doesn’t value efficiency because it sees research as weakness. Hesitation is framed as incompetence. Caution is cowardice. Thinking is “analysis paralysis.” Meanwhile, action, even dumb, costly action, is viewed as courage. But no one ever seems to calculate the cost of burning time and credibility on preventable errors. Probably because doing that would require… research.

Postscript

A couple small edits were made for spelling shortly after publication, because I didn’t delay gratification.

Larger Remark

But here’s a larger remark, the Stanford marshmallow experiment shows what I need it to show to use as an example, even if replication of the study suggests delaying gratification is tied to socioeconomic and other contextual factors, and those broad influences are driving the academic outcomes cited by the original study.

You, academics, and I can all theorize and comb over data to figure out why people will delay gratification or not. But the Stanford marshmallow experiment at face value asks the question; “ do you want one of something now, or two in 15 minutes?”

My core argument is simply that delaying gratification when a guarantee or near guaranty of an increased reward exist is a positive trait.

I suspect some of the reasons people involved in the Donald J Trump executive branch are unwilling to delay gratification despite being largely incredibly wealthy comes down to a few simple reasons.

1) The goals they say they have, are not the goals they are really working towards.

To be very clear if the goal was to cut costs for the US federal government DOGE is doing a poor job. The nonpartisan nonprofit, Partnership for Public Service (PSP) estimates that DOGE's actions will cost $135 billion this fiscal year. That cost doesn’t include the money spent fighting lawsuits challenging DOGE's actions.

I think the goal of DOGE, and other actions taken by leaders within Trump’s executive branch have been to punish people who have in the past oppose Trump or his major financial supporters, and to otherwise create confusion among those who would serve as opposition to Trump’s agenda in the future.

2) The consequences of being wrong do not impact the people taking the actions.

When the Trump administration fires or arrests someone, it’s a bad day for the person fired or arrested, not for Trump or his lieutenants. A small army of attorneys in the government will respond to lawsuits, appeal when needed, and rarely if ever will Trump or his most senior team have to even testify.

3) Trump rewards action

Trump seems to like flattery, and seeing manic flurries of activity. He is making the calls right now, and he would look at someone who said, “In 3 months we will have a list of people to fire, with reasons, without disruptions to core government services, and with very low risk of litigation,” and Trump would probably say his iconic phrase.