Cave team · Updated July 23, 2026 · 14 min read

How do you stop overthinking a decision?

TL;DR: Overthinking is a stopping problem, not a thinking problem. In research, chronic overthinkers made plans about as good as everyone else's — they just couldn't commit. More analysis buys confidence, not accuracy. What helps: set a "good enough" bar before comparing, lay the options out where they stay put, and give the decision a deadline. Cave, a private AI companion, helps you settle on a decision by talking it out.

If you've reopened the same pros-and-cons list for the third night running, you already suspect the extra thinking isn't working. The research below says your suspicion is right — and that's better news than it sounds, because it means the way out isn't a smarter analysis. Here's what's actually going on when a decision won't close, and what reliably closes it.

Why can't I make a decision?

Probably because you're ruminating, not deliberating — re-deciding a decision that is, by outside standards, already made. That distinction comes from a study designed to catch it. Psychologists Ward, Lyubomirsky, Sousa and Nolen-Hoeksema asked habitual ruminators — people who chronically chew on their problems — and people who rarely ruminate to each devise a plan, then had the plans rated (2003). The ruminators' plans came out only slightly lower: 4.43 versus 4.63. The real difference appeared after the planning was done. Asked whether they were ready to commit, 28.6% of the ruminators said they'd need more time before feeling comfortable — versus 1.4% of everyone else. The authors call it "behavioral paralysis," and it's the most useful reframe in this literature: the plan you already have is probably fine. What's failing is the stopping.

The obvious objection: maybe the extra laps still buy a better decision, so the lost sleep pays for itself. Two classic experiments tested exactly that — one on thinking harder, one on gathering more information.

Doesn't more analysis make the choice better?

Past an early point, no. In experiments, thinking harder made choices worse, and gathering more information made people confident without making them right.

Thinking harder first. Tim Wilson and Jonathan Schooler had students taste and rank five strawberry jams that Consumer Reports' trained panel had already ranked (JPSP, 1991). Students who simply tasted agreed with the experts closely — a .55 correlation. Students told to first analyze why they felt the way they did almost stopped agreeing with the experts: their correlation fell to .11, statistically no better than chance. Writing out plausible-sounding reasons drowned out what they could plainly taste. A follow-up found the reason-analyzers were also less satisfied with their choices weeks later.

Now the second experiment: gathering more information. In 1965, psychologist Stuart Oskamp gave 32 clinical judges one person's real case history in four installments, each adding more background, and had them answer the same questions about the person after every installment. The extra material barely moved their accuracy: it started at 26% and never reached 30%. Their confidence rose anyway — from 33% to 53% — so by the final round, more than 90% of them were surer than their answers justified. That's what another hour of research tonight buys: certainty, not correctness.

Confidence rises with more information; accuracy does not Two lines across four information stages from Oskamp 1965. Confidence climbs from 33 percent to 53 percent. Accuracy starts at 26 percent and stays below 30 percent. Over 90 percent of judges were overconfident by the final stage. More information bought certainty, not correctness 32 judges reading a case in four installments · Oskamp 1965 stage 1 stage 2 stage 3 stage 4 confidence 33% 53% accuracy 26% still under 30%
The same result as a picture: 32 clinical judges answered the same questions four times, with more information in hand each time. Accuracy (blue) went flat almost immediately; confidence (amber) kept climbing — by the final stage, over 90% of the judges were more confident than their accuracy justified. Oskamp, Journal of Consulting Psychology, 1965.

So the loop doesn't earn its keep. These are two experiments about two particular kinds of decisions — one a matter of taste, one of expert prediction — and the pattern recurs in decisions like those; it isn't a universal law. The next question is why this particular decision started a loop.

Why does this one decision have you stuck?

Usually because the options are hard to compare — not because there are too many of them. The "too many options" story is the famous one — and as it happens, it's also about jam. A supermarket tasting stand offering 24 jams drew more shoppers than one offering six, but buying ran the other way: 30% of the shoppers who saw six bought a jar, versus 3% at the big display. That image went everywhere. The effect mostly didn't survive testing: researchers later pooled every experiment they could find that tested it — 5,036 people choosing among jams, chocolates, pens, vacation destinations — and found that adding options made, on average, virtually no difference (Scheibehenne et al., Journal of Consumer Research, 2010).

What survived is more useful. A larger analysis — 99 results, 7,202 people — found that choice paralysis reliably appears under four specific conditions: the options are hard to compare, the decision is difficult, you have no clear prior preference, and you want the choosing over with (Chernev et al., 2015). That's a diagnostic checklist. Two job offers paralyze you not because two is many, but because they're incomparable — salary against meaning — the stakes are high, and you've never had to know which of those you value more. The fix isn't shrinking the list. It's making the options comparable, and that's a values question, not a research question.

Are you a maximizer? The personality of overthinking

Some people overthink one decision. Some overthink every decision, because their rule is "keep searching until I've found the best." Barry Schwartz's studies of 1,747 people gave that style a name — maximizing — and a warning: maximizers score higher on regret (r = .52, a strong link), higher on depression (r = .34, a moderate one), and lower on happiness than satisficers, who set a bar in advance and take the first option that clears it (JPSP, 2002).

The sharpest test tracked 548 graduating students through their job hunts (Iyengar, Wells & Schwartz, Psychological Science, 2006). Maximizers won the objective game: starting salaries of $44,515 versus $37,085 — 20% higher. They were also less satisfied with the job they took — 7.02 versus 7.58 on a 9-point scale — and felt worse day to day while searching, and the dissatisfaction held even after controlling for the higher salary. The exhaustive search worked. It cost more than it paid.

Maximizers earned more and liked their jobs less Two bar panels comparing maximizers and satisficers among 548 job-seeking students. Left panel, average starting salary: maximizers 44,515 dollars, satisficers 37,085 dollars — about 20 percent higher. Right panel, satisfaction with the accepted job on a one-to-nine scale: maximizers 7.02, satisficers 7.58. The exhaustive searchers earned more — and liked their jobs less 548 graduating students tracked through their job hunts · Iyengar, Wells & Schwartz 2006 Average starting salary $44,515 (+20%) $37,085 maximizers satisficers Satisfaction with the job taken (1–9) 7.02 7.58 maximizers satisficers
Both findings, same 548 students. Left: average starting salary — maximizers earned about 20% more. Right: satisfaction with the accepted job on its 1–9 scale — maximizers scored lower. Dollars on the left, a rating on the right, so compare bars within a panel, not across panels. Iyengar, Wells & Schwartz, Psychological Science, 2006.

"Satisficing" isn't settling, by the way. Herbert Simon — who later won the Nobel Prize for the idea — observed that real minds "satisfice; they do not, in general, optimize" (1956). Good-enough-by-a-defined-bar is the decision rule brains are built for. "The best possible" is the one that never terminates.

Isn't it just decision fatigue?

Decision fatigue is on much shakier ground than the internet suggests. Its star evidence is the "hungry judges" study: Israeli parole approvals appeared to fall from about 65% to near zero as court sessions wore on, recovering after food breaks (Danziger et al., PNAS, 2011). Three problems turned up. Case order wasn't random — a re-analysis found prisoners without lawyers tended to be scheduled last and won parole far less often, 15% versus 35% (Weinshall-Margel & Shapard, 2011), though the original team replied that the drop survived controlling for representation. The effect is far too large to be fatigue — about ten times the size of typical mental-fatigue effects (Glöckner, 2016). And the theory underneath — ego depletion, the idea that willpower drains like a fuel tank — failed its retest: 23 labs re-ran it on 2,141 people under preregistered conditions and measured d = 0.04, indistinguishable from zero, where the older literature had claimed a large effect, d = 0.62 (Hagger et al., 2016).

Tiredness is real, and a hard day makes every choice feel heavier. But "my willpower tank is empty" explains the midnight loop worse than the mechanism this post opened with: rumination. A loop wants an exit, not a nap.

What actually helps, ranked by evidence?

Externalize, pre-commit, and cap the search. The fixes with real evidence behind them, strongest first:

  1. Get the options out of your head and onto a surface that keeps them. The best-tested version is the humble decision aid — a structured layout of the options and what you value. Across 209 randomized trials with 107,698 people facing real medical choices, decision aids improved knowledge, cut "feeling uninformed" and "unclear about my values," and did not increase decision regret (Cochrane, 2024). Working memory is where options blur together; a page is where they hold still.
  2. Give the decision a deadline it doesn't have to re-win. "If it's Sunday at 6 p.m., I send my answer." Psychologists call these if-then plans implementation intentions; across 94 tests they closed the gap between deciding and doing with a medium-to-large effect (d = .65) (Gollwitzer & Sheeran, 2006). The decision stops being a nightly referendum.
  3. Set the "good enough" bar before you compare. A salary floor, a commute cap, one non-negotiable — then take the first option that clears them. That's satisficing by design; the maximizer numbers above are the running cost of the alternative policy.
  4. Make the incomparable comparable. Per the four-condition checklist, a stuck decision is usually missing a values ranking, not another spreadsheet column. "Which matters more this decade — money or meaning?" is answerable; "which option is best?" isn't until you've answered it. Talking the options through surfaces that ranking faster than staring at the list does.
  5. If it's a genuine tie, treat the tie as the answer. In a field experiment on real decisions, people on the fence who were nudged toward change by a coin flip were significantly happier six months later (Levitt, 2021 — participants were self-selected, so read it as evidence about ties, not a command to leap). If a coin could settle it, both options are survivable.

And one guard: if the loop keeps running at night after the decision is effectively made, that's no longer deliberation — dump it before bed and give it an appointment tomorrow. If rumination is a constant across your whole life rather than one hard month, that pattern is worth taking to a professional, not an app.

Where does the deliberating actually stop?

In a place where the options and your reasons are laid out and stay put. That's what the strongest evidence above has in common: the intervention that beat paralysis across 209 trials was, at bottom, a structured external surface for options and values. Cave is built to be that surface in conversation form — a private space to think out loud with an AI companion that keeps what you say as memory you can open and reread, organized by topic. You talk the decision through with Flamy, who asks the questions a spreadsheet can't ("which of these would you protect if you could only keep one?"), and the reasoning stays: Tuesday's clarity is still on the page on Thursday, connected to what you said about the same decision three weeks ago. Flamy won't pick for you. What changes is that the fourth re-decision of the same decision gets recognized as rumination — and the deliberating finally settles into a decision instead of looping.

Sources

FAQ

How do you stop overthinking a decision?

Change the stopping rule, not the amount of thinking. Set a "good enough" bar before comparing, lay the options out on paper or in one conversation so they stay put, and give the decision a deadline with an if-then plan ("Sunday, 6 p.m., I answer") — a method that improved follow-through across 94 tests. Past that point, more research mostly raises confidence, not accuracy.

Is analysis paralysis a real, documented thing?

The paralysis is real; the popular explanation isn't quite. Adding options — jams, chocolates, pens — averaged out to no effect across 63 experimental conditions. Paralysis reliably appears only when options are hard to compare, the decision is difficult, you have no clear preference, and you want the choosing over with. The measured mechanism looks like rumination: chronic ruminators produced plans rated about as good as anyone's, but 28.6% couldn't commit to them, versus 1.4% of everyone else.

What is a maximizer versus a satisficer?

A maximizer keeps searching until they've found the best possible option; a satisficer defines "good enough" in advance and takes the first option that clears it. In a study tracking 548 job-seeking students, maximizers landed 20% higher starting salaries ($44,515 vs $37,085) yet liked the jobs they took less and felt worse during the search. Maximizing also correlates with regret (r = .52) and lower happiness.

Is decision fatigue actually real?

The famous version is shaky. The "hungry judges" study has a documented scheduling confound, an effect about ten times too large for fatigue, and its underlying theory — ego depletion — produced a near-zero result (d = 0.04) when 23 labs re-ran it on 2,141 people. Being tired matters, but midnight paralysis is better explained by rumination — a loop that needs an exit, not a nap.

What's a practical way to stop re-deciding the same decision every night?

Give the deliberation one place to live. Talk or write the options through once, properly — values, fears, the good-enough bar — and keep the record. On Cave, you talk the decision out with Flamy until you can settle on an answer, and the conversation stays as memory you can open and reread, organized by topic — so Thursday-you rereads Tuesday-you instead of starting the loop again. If the loop restarts anyway, treat that as rumination, not new information.