Distinguish steelmanning from sycophancy

Charitably interpreting a position is not the same as agreeing with it, refusing to disagree, or softening your critique to avoid conflict.

Why it works

Steelmanning requires intellectual courage: you must construct the strongest version of a view and then honestly engage with it — which means saying clearly when even the strongest version fails. Sycophancy produces the appearance of charitable engagement without the substance: the "steelman" is constructed to justify agreement, not to find the strongest challenge. The distinction is that a genuine steelman increases the probability of updating your own view; a sycophantic one is designed to preserve it.

How to do it

1. After constructing a steelman, ask: "Has this steelman actually changed anything I believe or added anything I hadn’t considered?"
2. If the answer is consistently no, you are likely constructing steelmans designed to fail.
3. If the steelman reveals a genuine problem with your position, revise before continuing.
4. Track whether your positions ever actually update after steelmanning exercises — that is the calibration signal.

Evidence

The sycophancy failure mode is documented in AI systems and is a recognized risk in human social reasoning — social pressure creates a systematic incentive to validate rather than challenge. The distinguishing criterion here (does the steelman produce genuine updates?) is a practitioner diagnostic rather than an experimentally validated technique. (anecdotal)

Deciding whether an update is "genuine" or motivated is itself subject to motivated reasoning. External accountability partners can help calibrate.

Common mistake

Concluding after each steelmanning exercise that your position was right all along, without tracking whether you ever actually update — a consistent pattern of no-update is strong evidence of sycophantic steelmanning.