This publication is produced by a firm that advises people for a living, which makes this article awkward and worth writing for exactly that reason. The literature says advice is systematically underweighted, that the reasons are not really about the advice, and that a reputation as an adviser is far easier to lose than to build.
Key Takeaway
Human judges exhibit egocentric advice discounting: they "favor their own opinion over the advice even when the advice is good"[1]. On reputation: "the reputation gain is slow (requires a lot of samples of good advice), while the reputation loss is fast (a good turned bad advisor is heavily discounted after just a couple of bad suggestions)"[1]. And regardless of whether advice is taken, "the confidence of the judge in their final decision goes up", while there are "conflicting reports on whether the accuracy of such a final decision is changed"[1].
Our Grades For These Claims
Applying the scheme from the first article in this series.
Grade B that people discount advice. It is described as one of the main findings of the literature by two independent peer-reviewed sources, with an integrative review and multiple primary papers behind it, though we obtained none of them in full.
Grade C for the reputation asymmetry, which reaches us through a single academic preprint's summary of the literature.
Grade C for the confidence finding, from the same source, which itself notes conflicting reports on the accuracy side.
Our position: the sourcing here is thinner than the topic deserves, and the article is built around what multiple independent sources agree on rather than any single claim.
A Note On Method
Everything here is verified to August 2026.
We obtained no paper in this article in full. The foundational 2000 paper has no abstract available on the bibliographic record we found[2], and the 2004 paper's abstract truncates before its results[3].
Our two substantive sources are a clinical study protocol summarising the literature[4] and an academic preprint's literature section[1]. Neither is a primary source and we flag both throughout.
We obtained one abstract in full, for a 2012 paper, but it truncates before its result[5].
Everything else is titles and citations from bibliographic databases.
All simulation is ours.
This article discusses research on advice. It is not professional conduct, client engagement or practice management advice.
The Judge And The Advisor
The paradigm, which has standard vocabulary worth knowing.
A preprint describes it: "the advisee is called a 'judge' and the person or a system providing additional information or suggestions is called an 'advisor'."[1] The framework is referred to in the literature as the Judge-Advisor System[6].
The 2004 paper describes the method: "Respondents were asked to provide final judgments on the basis of their initial opinions and advice presented to them. The respondents' weighting policies were inferred."[3]
Two observations, ours.
The design is elegant and quantitative. Because you know the initial opinion, the advice, and the final judgment, you can compute exactly what weight was placed on the advice. It is not a self-report.
And the same paper notes that until recently "little attention has been given to it in either empirical studies or theories of decision making"[3], despite advice-seeking being, in its words, a basic practice in making real life decisions.
The Finding
What the literature reports, from a peer-reviewed source summarising it.
A clinical study protocol reviewing this literature states: "One of the main findings reported in the literature is that people tend to discount advice. Although the appropriate use of advice leads to better judgements (because of the reduction of random error), respondents prefer their own estimates; this is referred to as egocentric discounting." It adds that "the distance of the advice from one's own opinion can affect the weight it receives."[4]
Three observations, ours.
The parenthesis matters: advice helps because of the reduction of random error. That is the mechanism from the previous article in this series, stated by a different literature.
Even when the advice is good is the phrase from our other source[1]. This is not a rational response to poor advice; the discounting persists when the advice is accurate.
And the distance of the advice from one's own opinion affects the weight. Advice that differs a lot is discounted more, which is exactly backwards: distant advice is the advice most likely to bracket the truth and therefore most valuable to incorporate.
What Discounting Costs
Putting a number on it. Our own simulation, with two judges of equal accuracy and independent errors, invented parameters, from no source.
Under those assumptions the accuracy-maximising weight on the advice is 50 percent. Here is the mean absolute error at other weights, indexed to the best.
At 0 percent, ignoring advice entirely: 41.1 percent more error.
At 10 percent: 28.0 percent more.
At 20 percent: 16.7 percent more.
At 30 percent: 7.8 percent more.
At 40 percent: 1.9 percent more.
Two observations.
These figures are illustrative only and depend heavily on the equal-accuracy assumption. If the adviser is genuinely worse, the optimal weight is below 50 percent and some discounting is correct.
But the direction holds under any reasonable assumption: the penalty for ignoring a competent second opinion is large, and larger than most people would guess.
The Shape Of The Penalty
The most practically useful feature of those numbers. Ours.
Look at where the cost sits. Moving from 0 to 20 percent weight recovers about 24 points of the 41 available. Moving from 40 to 50 recovers under 2.
Two consequences.
The curve is steep near zero and flat near the optimum, which means the important decision is whether to give advice any real weight at all. Precision about the exact weight is nearly worthless.
Which is good news practically. Nobody has to compute anything. Moving from dismissing a second opinion to taking it moderately seriously captures most of the available gain, and the remaining calibration is not worth arguing about.
Choosing Instead Of Blending
A separate error, and it connects directly to the previous article.
The clinical protocol records: "A surprising result of previous research has been that averaging was the more effective strategy across a wide range of commonly encountered environments. Authors have observed that despite this finding, choosing was the preferred strategy although greater accuracy would have been achieved had they always averaged."[4]
The reference is to Soll and Larrick (2009)[4], whose citation appears elsewhere as Strategies for revising judgment[7]. We did not obtain it.
Two observations, ours.
On our own simulation, choosing one estimate at random costs about 41 percent more error than averaging the two, which is the same penalty as ignoring advice entirely. That is not a coincidence; picking your own estimate is ignoring the advice.
And note that the previous article's author is one of these authors. The same research group established that averaging works, that people do not believe it works, and that when given the chance they choose rather than average.
The Reputation Asymmetry
The finding with the sharpest commercial edge, and the one that should concern any adviser.
A preprint summarising the literature states: "To assess the quality of the advisor who gives more than one piece of advice, humans continuously update reputation judgments: the reputation gain is slow (requires a lot of samples of good advice), while the reputation loss is fast (a good turned bad advisor is heavily discounted after just a couple of bad suggestions)."[1]
This reaches us through a single preprint's characterisation of the literature, not from a primary paper, and we grade it C accordingly. The foundational 2000 paper's title includes reputation formation[2], which is consistent, but we could not obtain its abstract.
Three observations, ours.
If accurate, an advisory relationship is structurally fragile in a way that is not proportional to performance. Years of correct calls build slowly; two wrong ones undo them quickly.
This is the same shape as the twenty-third article in this series, which found that integrity violations move a judgment far more than competence violations because a single incident is highly diagnostic. Here the asymmetry appears in competence itself.
And it explains something advisers experience and rarely name: the client who leaves after one bad year was not being unreasonable by their own decision rule. They were applying an update process that weights bad evidence more heavily, which the literature describes as normal.
What That Means For An Adviser
Our own reasoning, and it is uncomfortable for the firm publishing it.
Three consequences if the asymmetry holds.
Recording the basis of advice matters more than being right on average. If two bad outcomes can undo a long record, the defence is a contemporaneous account of why the advice was reasonable when given, which distinguishes a bad outcome from bad advice.
Calibrated language is protective. Advice given with an honest confidence level survives a bad outcome better than advice given flatly, because the outcome was inside the stated range. The sixteenth article in this series argued calibration is the scoreable property of a track record.
And the incentive runs toward conservatism, which is worth naming. If the penalty for being wrong exceeds the reward for being right, an adviser optimising for reputation will under-recommend action. That is a real distortion and it is not in the client's interest.
The Confidence Problem
The finding we find most troubling.
From the same preprint: "No matter if the advice is taken or not, the confidence of the judge in their final decision goes up, compared to the pre-advice condition." And: "surprisingly, there are conflicting reports on whether the accuracy of such a final decision is changed by receiving advice."[1]
Three observations, ours.
Read those together. Confidence rises reliably; accuracy may not. That is a calibration failure, and it is produced by the act of consulting rather than by the content of what was said.
It means seeking advice can make a person worse off in one specific sense: more certain of a judgment that has not improved, and therefore less likely to revisit it.
And it applies to the person who ignored the advice as well. Having consulted and dismissed, they now hold their original view with more conviction than before, which is close to the opposite of what consultation is for.
The Bad Thing About Good Advice
A paper whose title states the problem, recorded because it names it precisely.
A bibliographic record identifies Soll, J. B., Palley, A. B., and Rader, C. A. (2022), The Bad Thing About Good Advice: Understanding When and How Advice Exacerbates Overconfidence, Management Science, 68(4), 2949–2969[7].
We did not obtain this paper and report its title only.
Two observations, ours.
The title asserts that advice exacerbates overconfidence, which is a stronger claim than the confidence finding above and is consistent with it.
And the qualifier good advice is the point. The paper is not about being misled. It appears to concern a cost that arrives with advice that is correct, which is a genuinely counterintuitive proposition and one we cannot report the substance of.
What Changes The Discount
The conditions, from two sources.
A preprint states that "the bigger the perceived expertise of the advisor is, the lesser the discounting", and that "people follow extreme advice less and average more frequently when they are not sure if they or the advisor are more knowledgeable about the domain"[1].
Another states that "discounting is more pronounced when decision makers have less prior knowledge, face lower uncertainty, or operate in low-stakes tasks"[6].
Three observations, ours.
The expertise effect is unsurprising and useful: perceived expertise reduces discounting, which is an argument for making the basis of expertise visible rather than assuming it is known.
The claim that discounting is more pronounced when decision makers have less prior knowledge is counterintuitive, since less knowledge should mean more reason to defer. We report it as the source states it and flag that we did not obtain the underlying work.
And low stakes producing more discounting is practically encouraging: it implies that on the decisions that matter most, advice gets more of a hearing.
Does Paying For It Help
The question a fee-charging firm has to ask, and our gap on it.
A bibliographic record identifies Gino, F. (2008), Do we listen to advice just because we paid for it? The impact of advice cost on its use, in Organizational Behavior and Human Decision Processes[8]. A related paper is identified as Godek, J., and Murray, K. B. (2008), Willingness to pay for advice: The role of rational and experiential processing, OBHDP, 106(1), 77–87[8].
We did not obtain either and report no findings from them.
Two observations, ours.
The title is a question and we cannot answer it. That is a real gap, sitting exactly where a professional firm's interest lies, and we would rather say so than infer.
We note only that the question cuts both ways. If paid advice is weighted more heavily, that is a commercial argument for charging. If it is weighted more heavily regardless of quality, that is a warning about what a fee actually buys in a client's mind.
Algorithmic Advice Is Discounted Less
A finding relevant to every firm now deploying software that recommends things.
A preprint reports: "the literature describes evidence of egocentric advice discounting in algorithmic advice but points out that users discount algorithmic advice to a lesser extent than advice from individuals." And: "Humans are more inclined to accept suggestions from an algorithm that has been proven correct in the past."[1]
We did not obtain the underlying work.
Two observations, ours.
If accurate, this is the reverse of algorithm aversion as usually described, and this publication has covered the aversion literature elsewhere. The two findings are not necessarily contradictory, since one concerns choosing whether to use an algorithm and the other concerns how much weight its output receives, but the tension is worth flagging rather than smoothing over.
And the reputation sentence applies to software as much as to people. An automated recommendation that has been wrong twice will be discounted heavily, on the same asymmetry.
The Blindfold Effect
An intervention with a memorable name, and a truncated result.
Yaniv and Choshen-Hillel published Exploiting the Wisdom of Others to Make Better Decisions: Suspending Judgment Reduces Egocentrism and Increases Accuracy in the Journal of Behavioral Decision Making, 25, 427–434, in 2012[5].
Their framing: "Although decision makers often consult other people's opinions to improve their decisions, they fail to do so optimally. One main obstacle to incorporating others' opinions efficiently is one's own opinion. We theorize that decision makers could improve their performance by suspending their own judgment."[5]
The design: in three studies estimating caloric values, "in the full-view condition, participants could form independent estimates prior to receiving others' opinions, whereas participants in the blindfold condition could not form prior opinions. We obtained an intriguing blindfold effect."[5]
Our source truncates before stating what the effect was. The title asserts that suspending judgment reduces egocentrism and increases accuracy, and we report that as the title's claim rather than as a result we have read.
A Tension With The Previous Article
Something we have to address rather than leave for a reader to notice. Ours.
The thirty-sixth article argued that independence is essential: get the second opinion before the first is heard, because a second person who has seen the first number is not a second judgment.
This article reports work arguing that your own prior opinion is the main obstacle to using someone else's, and that not forming one first may help.
Two ways to reconcile these, ours and offered as reasoning rather than resolution.
They concern different roles. The person aggregating several judgments needs those judgments to be independent of each other. The person receiving advice is a different case, because they will overweight whichever opinion is theirs.
So the design that satisfies both is: collect independent judgments, and have them combined by someone who did not produce one. That removes the egocentric weight from the combination step while preserving independence at the input step.
We flag that this reconciliation is our own construction and is not proposed by any source we obtained.
What An Advisory Firm Can Do
The practical translation, written by a firm that gives advice. Ours, untested.
Four observations.
Expect to be discounted, and do not read it as a failure of the relationship. The literature describes this as general, persisting even when advice is good.
Ask what the client already thinks, and when. Advice given after a client has formed a firm view competes with that view. Advice sought before is in a different position, and the blindfold work suggests the difference is not small.
Do not soften advice toward the client's position to get it accepted. Distant advice is discounted more, so there is a real temptation to move closer. But the value of a second opinion comes precisely from its independence, and advice adjusted to be palatable has had its usefulness removed to make it acceptable.
And record the reasoning, not just the recommendation. If reputation loss is fast, the only durable protection is a contemporaneous record showing the advice was sound when given.
What To Do
Assume you are discounting good advice. The literature reports people favouring their own opinion even when the advice is good, and this is described as a main finding rather than an edge case.
Give advice real weight rather than precise weight. On our simulation the penalty curve is steep near zero and flat near the optimum, so moving from dismissal to serious consideration captures most of the gain.
Be most suspicious of your reaction to distant advice. Advice far from your view is discounted more and is exactly the advice most likely to be worth incorporating.
Blend rather than choose. Picking one estimate costs about as much as ignoring the advice entirely, because it is the same thing.
Notice that confidence rises either way. Consulting raises certainty whether or not the advice is used, while the accuracy effect is reported as contested. Feeling more sure after asking someone is not evidence you are more right.
If you advise, record why. Reputation gain is reported as slow and loss as fast, so a contemporaneous account of the reasoning is what distinguishes a bad outcome from bad advice.
Do not move your advice toward the client to get it accepted. That trades away the independence that made it valuable.
Separate the roles. Have judgments collected independently and combined by someone who did not produce one, which is our own suggestion rather than a finding.
The Limits Of This Analysis
Several caveats matter. This article discusses research on advice and is not professional conduct, client engagement or practice management advice. Everything is verified to August 2026. We obtained no paper in this article in full. The foundational 2000 paper has no abstract available on the bibliographic record we located; the 2004 paper's abstract truncates before its results; the 2012 paper's abstract truncates before stating the effect it names. Our two substantive sources are a clinical study protocol and an academic preprint, both summarising the literature rather than reporting it, and we flag them at every use. The reputation asymmetry and the confidence finding each rest on a single preprint's characterisation and are graded C accordingly. We did not obtain the papers on advice cost, which sit exactly where a fee-charging firm's interest lies, and report no findings from them. We did not obtain the 2022 paper on advice and overconfidence and report its title only. All simulation is ours, uses invented parameters, and assumes two judges of equal accuracy with independent errors; if an adviser is genuinely less accurate, some discounting is correct and our figures overstate the penalty. The reconciliation offered between this article and the previous one is our own construction, proposed by no source. The observation about advisers' incentives running toward conservatism, and all practical sections, are our own reasoning.
Frequently Asked Questions
What is egocentric discounting?
How much does it cost?
Do I need to get the weighting exactly right?
Why do advisers lose clients over one bad year?
Does asking for advice at least make me more confident correctly?
Doesn't this contradict the previous article?
References
- Academic preprint on AI fact-checking and authority, literature section, on the advisee being called a judge and the person or system providing suggestions an advisor; on human judges exhibiting egocentric advice discounting, favoring their own opinion over the advice even when the advice is good; on the bigger the perceived expertise of the advisor being, the lesser the discounting; on humans continuously updating reputation judgments, with the reputation gain being slow and requiring a lot of samples of good advice while the reputation loss is fast, a good turned bad advisor being heavily discounted after just a couple of bad suggestions; on people following extreme advice less and averaging more frequently when unsure whether they or the advisor are more knowledgeable; on the confidence of the judge in their final decision going up whether or not the advice is taken, compared to the pre-advice condition; on there being conflicting reports on whether the accuracy of such a final decision is changed by receiving advice; on evidence of egocentric advice discounting in algorithmic advice, with users discounting algorithmic advice to a lesser extent than advice from individuals; and on humans being more inclined to accept suggestions from an algorithm proven correct in the past. Note: an academic preprint's summary of the literature, not a primary source. The reputation asymmetry and the confidence finding rest on this source alone and are graded C accordingly. arxiv.org
- Bibliographic record for Yaniv, Ilan, & Kleinberger, Eli (2000), Advice Taking in Decision Making: Egocentric Discounting and Reputation Formation, Organizational Behavior and Human Decision Processes, 83(2), 260–281, November. Note: this record states that no abstract is available for the item, and full text access is restricted. We report the citation and title only; the title's reference to reputation formation is consistent with the claim at reference 1 but does not establish it. ideas.repec.org
- Working paper record for Yaniv, Ilan, Receiving Other People's Advice: Influence and Benefit, published as Organizational Behavior and Human Decision Processes, 93(1), 1–13, January 2004, on seeking advice being a basic practice in making real life decisions to which little attention had until recently been given in either empirical studies or theories of decision making; on the studies investigating the influence of advice on judgment and the consequences of advice use for judgment accuracy; and on respondents having been asked to provide final judgments on the basis of their initial opinions and advice presented to them, with the respondents' weighting policies inferred. Note: a working paper record; the abstract truncates before reporting any results and we report none. ideas.repec.org
- Clinical study protocol reviewing the advice-taking literature, on one of the main findings being that people tend to discount advice, citing Yaniv (2004), Yaniv and Kleinberger (2000) and Bonaccio and Dalal (2006); on the appropriate use of advice leading to better judgements because of the reduction of random error while respondents prefer their own estimates, referred to as egocentric discounting; on the distance of the advice from one's own opinion affecting the weight it receives; and on a surprising result of previous research being that averaging was the more effective strategy across a wide range of commonly encountered environments, with authors observing that choosing was nonetheless the preferred strategy although greater accuracy would have been achieved had they always averaged, citing Soll and Larrick (2009). Note: a clinical study protocol summarising the literature, not a primary source. cdn.clinicaltrials.gov
- Yaniv, I., & Choshen-Hillel, S. (2012). Exploiting the Wisdom of Others to Make Better Decisions: Suspending Judgment Reduces Egocentrism and Increases Accuracy. Journal of Behavioral Decision Making, 25, 427–434. DOI 10.1002/bdm.740, publisher record, on decision makers often consulting other people's opinions to improve their decisions but failing to do so optimally; on one main obstacle to incorporating others' opinions efficiently being one's own opinion; on the authors theorising that decision makers could improve their performance by suspending their own judgment; on three studies in which participants used others' opinions to estimate uncertain quantities, being the caloric value of foods; on the full-view condition allowing participants to form independent estimates prior to receiving others' opinions while the blindfold condition did not; and on the authors having obtained an intriguing blindfold effect. Note: a publisher record; the abstract truncates before stating what the blindfold effect was. The claim that suspending judgment increases accuracy is reported here as the paper's title, not as a result we have read. onlinelibrary.wiley.com
- Academic preprint on strategic advice, literature section, on a large behavioral literature centred on the Judge-Advisor System documenting systematic patterns in advice aggregation including egocentric discounting and its variation across contexts, with discounting more pronounced when decision makers have less prior knowledge, face lower uncertainty, or operate in low-stakes tasks; and on that literature providing experimental support for modeling advice aggregation as linear weighted averaging. Note: an academic preprint's summary, not a primary source. The claim about less prior knowledge is counterintuitive and we did not obtain the underlying work. arxiv.org
- Bibliographic citation record listing works citing Yaniv and Kleinberger (2000), including Soll, Jack B., Palley, Asa B., & Rader, Christina A. (2022), The Bad Thing About Good Advice: Understanding When and How Advice Exacerbates Overconfidence, Management Science, 68(4), 2949–2969; Soll, Jack B., & Mannes, Albert E. (2011), Judgmental aggregation strategies depend on whether the self is involved, International Journal of Forecasting, 27(1), 81–102; Ronayne, David, & Sgroi, Daniel (2018), When Good Advice is Ignored: The Role of Envy and Stubbornness; and Atanasov, P., Witkowski, J., Ungar, L., Mellers, B., & Tetlock, P. (2020), Small steps to accuracy: Incremental belief updaters are better forecasters, Organizational Behavior and Human Decision Processes, 160. Note: citations only. We obtained none of these papers and report the 2022 title without its substance. ideas.repec.org
- Bibliographic record for Yaniv, Ilan, & Milyavsky, Maxim (2007), Using advice from multiple sources to revise and improve judgments, Organizational Behavior and Human Decision Processes, 103(1), 104–120, whose reference and citation lists identify Gino, Francesca (2008), Do we listen to advice just because we paid for it? The impact of advice cost on its use, Organizational Behavior and Human Decision Processes; Godek, John, & Murray, Kyle B. (2008), Willingness to pay for advice: The role of rational and experiential processing, Organizational Behavior and Human Decision Processes, 106(1), 77–87; Sniezek, Janet A., & Buckley, Timothy (1995), Cueing and Cognitive Conflict in Judge-Advisor Decision Making, Organizational Behavior and Human Decision Processes, 62(2), 159–174; and Jodlbauer, Barbara, & Jonas, Eva (2011), Forecasting clients' reactions: How does the perception of strategic behavior influence the acceptance of advice?, International Journal of Forecasting, 27(1), 121–133. Note: citations only. We did not obtain the papers on advice cost, which bear directly on a fee-charging firm, and report no findings from them. ideas.repec.org
This article discusses research on advice and is not professional conduct, client engagement or practice management advice. No paper discussed was obtained in full; the foundational paper's abstract was unavailable and two others truncate before their results. The two substantive sources summarise the literature rather than reporting it. All simulation is the authors' own and assumes two judges of equal accuracy with independent errors, under which some discounting would be incorrect; where an adviser is genuinely less accurate, some discounting is correct.