Using Svadhyaya to enhance Superforecasting

What each one is actually for

Svadhyaya

Svadhyaya is the fourth of the five niyamas in classical yoga philosophy — the personal observances, distinct from the more externally-focused yamas — and it means, most literally, one's own study. In practice it is the deliberate examination of your own recurring mental and emotional patterns: noticing, across many instances rather than just once, that you consistently go quiet in a certain kind of conflict, or that a specific category of comment reliably triggers defensiveness, or that a certain thought keeps recurring in slightly different clothes. Traditionally this was done partly through studying sacred or philosophical texts and holding your own reactions up against them as a kind of mirror, using the text's questions to notice what you'd otherwise scroll past in yourself. The point is not self-improvement in a narrow sense; it's making a pattern that has been running you implicitly, unnoticed, into something you can actually observe and, from there, choose about.

Superforecasting

Calibration training takes a different kind of unexamined impression — not about yourself, but about the world — and does something structurally similar to it. Instead of saying "I think this project will probably finish on time," which commits you to nothing checkable, you say "I think there's a seventy percent chance this finishes by the deadline," a specific number that can later be scored as right or wrong. The practice, refined through Tetlock's research on what separates skilled forecasters from unskilled ones, involves making a real volume of these numeric predictions across many different questions, writing them down before the outcome is known, and then, once outcomes arrive, checking your track record: across every time you said seventy percent, did the thing actually happen about seventy percent of the time, or were you consistently overconfident, or underconfident, in a pattern you can now see and correct.

What you get out of pairing them

Svadhyaya buys you access to a blind spot you cannot see by any other route — the frame you're currently looking through is, by definition, invisible from inside it, which is why a pattern can run your behavior for years while you experience each instance as a fresh, unrelated event rather than as the same thing recurring. Calibration training buys you something adjacent but distinct: proof, not opinion, about how reliable your own judgment actually is, because a felt sense of confidence and an actually accurate prediction turn out to be only loosely related, and most people never find out how loosely because they never write the number down and check it later. Put together, you get a way to apply that same proof-instead-of-opinion standard to the interior patterns svadhyaya surfaces — not just noticing "I tend to go quiet in conflict," a genuine but still vague observation, but eventually being able to predict, with a number, whether you will, in a specific upcoming conversation, and finding out afterward whether you were right, which is a sharper, more corrigible form of the same self-knowledge.

What Svadhyaya rests on

• Externalizing makes pattern visible randomised trials
  Putting an internal pattern into an external, structured form makes it visible and available for examination instead of remaining implicit and merely enacted.
• Decentering from thought randomised trials
  Shifting attention from the content of a thought to observing it as a passing event in awareness creates distance from that content instead of fusion with it.
• Self distancing improves judgment randomised trials
  Evaluating a situation from a distanced, third-person vantage point instead of an immersed first-person one improves the quality of judgment about it.

What Superforecasting rests on

• Decomposition improves estimate accuracy observational studies
  Breaking a complex uncertain question into smaller more tractable sub-questions produces more accurate estimates than judging the whole question at once.
• Explicit probability estimates improve calibration randomised trials
  Forcing a numeric probability rather than a vague verbal hedge makes a prediction falsifiable and trainable.
• Outcome feedback corrects judgment bias randomised trials
  Regularly scoring predictions against actual outcomes and reviewing the misses corrects systematic biases in judgment over time.

Where they actually meet

Both practices refuse to let an impression stay an impression. Svadhyaya takes a hazy, half-noticed inner tendency — something like "I think I get defensive sometimes when my work is questioned" — and insists on actually watching for it, instance by instance, until the vague sense sharpens into something specific enough to name. Calibration training takes an equally hazy outer impression — "this will probably work out" — and insists on the harder step of a number, then a later check against what actually happened. Different material, a feeling about yourself versus a guess about the world, but the identical refusal underneath: don't let the vague version stand in for the real one, when a more specific, checkable version is available if you're willing to do the work of stating it.

The principle underneath both

An impression that never gets written down or checked against anything stays exactly as vague as it started, no matter how many times you have it — recurrence alone doesn't sharpen a hunch. What actually sharpens it is stating it in a form specific enough to be wrong, and then finding out whether it was, which is a fundamentally different activity from simply noticing the impression again and nodding at its familiarity.

What that makes the real goal

Build a working feedback loop between what you believe — about a pattern in yourself, or about how a real-world question will resolve — and what's actually true, specific enough that the belief can be corrected by evidence instead of simply repeating itself indefinitely, unchanged, each time it recurs.

Other practices that serve the same goal

Structured 360-degree feedback processes pursue a version of svadhyaya's objective from the outside rather than the inside — instead of you trying to notice your own blind spot, which is difficult precisely because it's a blind spot, other people report what they actually observe, giving you external, if less controlled, data on the same question; the evidence here is observational rather than experimental, but it sidesteps the specific limitation of self-report. Prediction markets pursue calibration training's objective at a collective rather than individual level, aggregating many people's probability estimates into a single number that has shown, across a range of domains, real accuracy advantages over individual expert judgment, though the mechanism, aggregation rather than individual feedback, is different from calibration training's personal track-record approach. Pre-mortem analysis, where a team imagines a project has already failed and works backward to specific causes before it starts, targets a related blind-spot problem — overconfidence about your own plan — through imagined hindsight rather than either sacred-text reflection or numeric tracking, with modest but real evidence for reducing planning overconfidence.

A tradition that already does this

Benjamin Franklin's daily virtue-tracking practice, described in his own autobiography, already fused something close to this centuries before Tetlock's research existed: he kept a chart of thirteen named virtues and marked a small dot for each violation he noticed in himself each day, reviewing the chart weekly to see which virtues kept recurring as problems. It is, in effect, self-study turned into an explicit, trackable, checkable record rather than left as a private, unmeasured impression — the same fusion this pairing is describing, built independently, long before either yoga philosophy and probabilistic forecasting had any occasion to meet.

Using Svadhyaya to enhance Superforecasting

Calibration training is usually aimed entirely outward — will this election go a certain way, will this project ship on time — and rarely turned on the forecaster's own patterns as a subject worth forecasting. Borrow svadhyaya's attention and add a category of question calibration training doesn't usually include: predictions about your own reactions, not just external events. Before a specific, recurring kind of situation — a performance review, a difficult family call, a meeting where your work gets critiqued — write down an actual numeric estimate: what's the percentage chance you go quiet instead of speaking up, or get defensive instead of curious, in this specific instance. Then, afterward, score it honestly against what actually happened, the same way you'd score a forecast about anything else. This does something svadhyaya's traditional noticing alone often doesn't: it forces the vague sense of "I think I tend to do this" into a specific, falsifiable claim, and after enough instances, you get an actual calibration curve for your own behavior — evidence, not impression, about exactly how reliable that self-pattern really is and under precisely which conditions it shows up.

Using Superforecasting to enhance Svadhyaya

Self-study can stay diffuse if the only tool is open-ended noticing — you observe a pattern, feel a flicker of recognition, and move on, without the pattern ever getting specific enough to actually work with. Bring calibration training's core discipline, decomposition, into the practice directly: instead of a broad question like "how do I handle conflict," break it into the smaller, more specific sub-questions the way a forecaster breaks a big uncertain event into tractable pieces — what happens in the first ten seconds after I'm criticized, specifically; does my voice change before or after I notice I'm upset; is the pattern different with family than with colleagues. Answering each smaller question is more tractable than trying to characterize the whole pattern at once, and it mirrors exactly the decomposition move that improves forecasting accuracy on external questions. Once you have specific sub-patterns named this precisely, you can apply outcome feedback to them the same way a forecaster reviews a track record: check, across a real number of instances over a few weeks, whether what you predicted about your own reaction in each sub-situation actually held, and let the misses correct the picture rather than assuming your first read of the pattern was already complete.

Writing down a number before the 3pm feedback call

It's 2:45, and in fifteen minutes your manager is going to walk through a piece of work you know has real problems in it. You've noticed before, in a general way, that you tend to get quiet and start agreeing too fast in these calls, even when you actually disagree with part of the feedback — a pattern you've named to yourself loosely a few times without ever pinning it down further. Today, instead of just noting the familiar dread, you write one line in a notebook: seventy percent chance I go quiet and agree within the first two minutes, rather than actually saying what I think about the parts I disagree with. You close the notebook and take the call. At minute one, your manager raises the exact point you expected to bristle at, and you feel the familiar pull to just nod. You catch it, barely, say one sentence of actual pushback, then go quiet again for the rest of the call. Afterward, you go back to the line: not quite the full quiet-and-agree pattern, but closer to it than not — call it a partial hit. You write a second line underneath: the pull to go quiet arrives inside the first ninety seconds, specifically right after the first piece of criticism, not gradually across the call. That's a sharper, more useable piece of information than "I get quiet in feedback calls" was an hour ago, and it came from writing the number down before the call, not from noticing the pattern again afterward the way you always had.

When this is worth reaching for

A tendency to go quiet and agree too fast in a certain kind of meeting has already been named loosely, more than once, without ever getting pinned down further. That loose, repeated hunch, not the earliest stage of simply starting to notice one's own reactions, is specific enough to actually test and sharpen.

Where this does not apply

Turning self-observation into a scored, numeric exercise too early, before someone has any comfortable practice simply noticing their own patterns without judgment, risks making self-study feel like a performance review of the self, which tends to increase anxiety and self-monitoring in an unhelpful way rather than genuine insight. And calibration training's decomposition habit, useful for a recurring behavioral pattern, is the wrong tool for a single, one-off emotional experience — grief, a specific loss, a moment that isn't a pattern at all but simply a thing that happened once; breaking a singular experience into forecastable sub-questions can flatten something that doesn't need to be predicted, only felt and understood.

What the evidence does and does not support

Explicit probability estimation and outcome-feedback scoring, the core mechanics of calibration training, have real experimental support from forecasting-tournament research, including Tetlock's own studies, showing that forecasters who train this way measurably improve their calibration over time. Decentering and self-distancing, the closest research-backed cousins to what svadhyaya is doing when it turns a thought into something observed rather than lived inside, also have randomized-trial support, mostly from mindfulness and self-distanced-reflection research rather than from any study of svadhyaya as a named practice specifically, which comes from a contemplative tradition rather than a laboratory. Nobody has tested applying numeric calibration scoring to self-observed personal patterns, or measured whether it actually improves self-knowledge beyond what traditional self-study alone produces; this page connects two separately evidenced practices by their shared underlying mechanism, not by a study of the combination itself.