Should You Trust Yourself When You Doubt Your Own Thinking?
A Whisper in the Exam Room
Maya is taking a math test. She works through the hardest problem and lands on an answer: 14. She is nearly certain. Then her friend leans over and whispers, “You look really pale. Are you sure your brain is working right? You might be too tired to think straight.”
Now Maya has a weird problem. She still thinks the answer is 14, but she also has a new piece of information: her own thinking machine might be glitching. Should she change her answer? Keep it? How should she even begin to decide? This tangle is what philosophers call epistemic self-doubt — doubt about whether your own beliefs are formed reliably.
At stake is something anyone who wants to think well must face. When you find evidence that you might be confused, sleep-deprived, or biased, what does that evidence demand of you? Should you stick to your guns, or should you back away from your own conclusion? The question is not just about tests. It pops up whenever a friend disagrees with you, when you realize you are upset, or when you read a study saying people in your situation make mistakes.
Philosophers have chased this puzzle in many directions. Let’s look at the main attempts to understand whether self-doubt can ever be rational — and what you should do about it.
The Anti-Expert: Why You Can’t Believe You’re Always Wrong
Imagine someone who is never right. Whenever she believes something, it turns out false. If she believes there is a cat in the yard, there is no cat. If she does not believe there is a cat, a cat is definitely there. The philosopher Roy Sorensen called such a person an anti-expert (Sorensen, 1988).
Could you ever rationally believe that you are an anti-expert? It seems not, once you bring in two simple ingredients: coherence (your beliefs must not blatantly contradict each other) and self-knowledge (you know what you believe). Suppose you believe a claim, call it q. Because you know your own mind, you also believe that you believe q. But the definition of anti-expertise says that if you believe q, q is false — which directly clashes with your belief that q is true. Your beliefs would be at war with themselves. You cannot coherently think both “I believe q” and “Whenever I believe something it is false” while knowing you believe q.
This result makes self-doubt look nearly impossible. Sorensen argued that a consistent person who knows her own beliefs simply cannot believe she is an anti-expert. Other philosophers, like Earl Conee and David Christensen, pushed back. They imagined situations where the evidence that you are incompetent is so overwhelming that believing you are an anti-expert might be the only honest response — even if it makes you incoherent. After all, sometimes the world hands you such crushing evidence that your mind deserves to be doubted completely.
Yet even a softer version of self-doubt gets tangled. Consider a modest thought: “I’m not perfect.” In precise terms, you might think there is at least a 5% chance that you are highly confident of something false, or highly unconfident of something true. While this sounds gentle, it turns out to be no easier to combine with perfect self-knowledge than full anti-expertise was. The numbers simply refuse to add up without contradiction.
The lesson from this first approach is painful: if you know your own beliefs sharply, you cannot coherently assign much weight to the idea that those beliefs are unreliable. Something has to give.
The Reflection Rule: Should Your Confidence Match Itself?
Perhaps we need a subtler rule. Enter the Reflection Principle, sometimes called Self‑Respect. The idea is simple: your credence — your degree of belief, from 0% to 100% — in a claim q, given that you now have credence x in q, should be exactly x. If you are 80% sure it will rain on condition that you are 80% sure it will rain, then your final number had better be 80%. It seems barely deniable. Why would you be more or less confident than you think you are?
Self-doubt smashes straight into this rule. When Maya hears she looks exhausted, she might think: “Given that I am nearly certain the answer is 14, my true confidence should maybe be lower — say 60%.” Philosophers call this state Refraction: P(q given my credence is x) is less than x. It captures the feeling that your current level of belief is not the right one. But if Self‑Respect is a rational requirement, then Refraction is irrational. You cannot both be a coherent thinker and disapprove of your own belief in this way.
Some defenders of the rule reply that the moment you realize your credence is too high, you should simply change it to the value you approve of, and then you will satisfy Self‑Respect again. The trouble is that changing your mind requires a reason. If you cannot even think “my current credence is inflated,” where does the instruction to revise come from? You seem to need a self-doubting thought in order to fix your thinking, but that very thought seems forbidden.
These puzzles led other philosophers to look for a rule that lets you compare yourself with an ideal thinker — someone whose judgment is unclouded by fatigue, bias, or confusion.
Higher-Order Evidence: The Pilot and the Altimeter
Imagine you are a pilot. You redo a fuel calculation and become certain you have enough fuel to fly fifty miles farther than planned. Then you glance at the altimeter and see you are at 10,500 feet. You remember that at this altitude hypoxia can set in — a condition where your thinking gets fuzzy without you noticing. Now you have higher-order evidence: evidence not about fuel directly, but about how reliable your own thinking is (Christensen, 2010).
Should you still believe you have enough fuel? The philosopher Richard Feldman argued there are two sensible paths. One is steadfast: you judge that your original calculation is still supported by the data, so you hold onto your belief in F (enough fuel) and also hold onto the belief that your evidence supports F. The other is conciliatory: you lower your confidence in F because the altimeter gives you reason to worry, and you accept that your evidence might not support F after all. In both of these, your first-order belief and your second-order belief match — you are at peace with yourself.
What Feldman ruled out was level-splitting, also called epistemic akrasia. That is the state where you believe F but also believe that your evidence does not support F. It would be like saying, “I believe the plane has enough fuel, but my total evidence doesn’t back that up.” Many philosophers find this deeply weird. It seems to disrespect your own evidence.
Yet several recent thinkers — Timothy Williamson, Brian Weatherson, and others — have argued that akrasia can sometimes be rational. Their thought is that information about your own fallibility is weaker than ordinary evidence against the content of your belief. If your reasoning was actually sound and your fuel really is enough, then sticking with F while also acknowledging that you might have messed up could be the only honest way to respond. You are simply tracking the real evidential relation, even while admitting you might have misjudged it.
This dispute remains wide open. The challenge is to craft a rule that respects higher-order evidence without leaving you helpless or contradicted.
Calibrating Your Brain Like a Thermometer
Another approach borrows an idea from science: calibration. Think of your confidence as a measuring instrument. A thermometer is well calibrated if, when it reads 30°C, the actual temperature tends to be 30°C. Your beliefs are well calibrated if, for every level of confidence you have, the world actually matches that confidence about the same proportion of the time. If you are 90% sure of answers on a quiz, you should be right about 90% of the time.
The philosopher Sherrilyn Roush (2009) proposed capturing this with a calibration curve: an objective probability that q is true given that your credence in q is x. You can gather evidence about your own calibration curve — from past test scores, feedback from teachers, or knowledge about your current level of fatigue. Then a rational thinker should set her final credence in q to match the objective probability indicated by her curve, not her initial hunch. This is expressed in a simple rule called Cal:
Your credence in q, given that your credence is x and the objective probability of q given that credence is y, should be y.
This means if you learn that when you are 90% confident under rotten conditions you are actually right only 60% of the time, you should adjust your confidence down to 60%. The rule comes directly from two commitments most philosophers already accept: probabilistic coherence and the idea that your beliefs should respect known objective chances (a rule called the Principal Principle).
On the calibration view, self-doubt is not an incoherent tangle. You can think: “I am 90% confident, but I have good evidence that when I am this confident in these circumstances I am only reliable about 60%.” Those beliefs don’t contradict one another — they just report a mismatch between your subjective feeling and the objective facts about your brain. The rule Cal then tells you exactly how to repair the mismatch, without forcing you to pretend the doubt never happened.
The calibration approach also explains something important: higher-order evidence does not always push your confidence downward. If you discover that you tend to be underconfident — maybe you always think you did worse on a test than you actually did — the very same rule might boost your confidence. Self-doubt becomes a tool you can use to correct yourself in any direction.
Why It Matters: Thinking Straight When You’re Leaning
You will never fly a plane through thin air or face an official test that declares you an anti-expert. But you will face situations where you get a signal that your thinking might be compromised. You stayed up too late and now you are puzzling over a tricky homework problem. A friend you respect thinks the opposite about who was at fault in an argument. Your science teacher shows the class a chart proving that people consistently overestimate how much they know about climate change.
In each case, you are presented with higher-order evidence — information about the reliability of your own mind. Should you discard your original verdict or stick with it? The philosophical tangle around self-doubt is not just a brain teaser. It is a map of the decision you face whenever you try to be intellectually honest.
The approaches we have seen don’t all agree, but they push toward a similar ideal: a kind of inner harmony where your confidence and your estimate of your own reliability walk in step. Whether that harmony comes from obeying Self‑Respect, from matching the imagined perfect thinker, or from calibrating your confidence like a thermometer, the demand is that you not ignore evidence about your own flaws. The dispute is about how deeply that evidence should trouble you and which part of your mental life gets the final say.
So next time a quiet worry about your own brain creeps in — tiredness, emotion, a hunch that you might be wrong — you can ask yourself not just “What should I believe?” but also “How much should I trust my own believing?” That second question, as philosophers have found, is surprisingly wild and far from settled.
Think about it
- If you believe your brain isn’t working well because you’re exhausted, should you stop trusting all your beliefs, or can you use some of them to figure out which ones are wrong?
- Imagine a person who tends to be overconfident in every situation. Could learning this fact ever help him become accurately confident, or would he automatically overestimate his new, lowered confidence too?
- When you disagree with a friend who is just as smart as you, should you always meet in the middle, or can it be rational for both of you to stick to your original views?
