Do we advise as one likes? The alignment bias in social advice giving

Xitong Luo; Lei Zhang; Yafeng Pan

doi:10.1371/journal.pcbi.1013732

Do we advise as one likes? The alignment bias in social advice giving

Xitong Luo
, Lei Zhang^*
, Yafeng Pan^*

^*Corresponding author for this work

Research output: Contribution to journal › Article › peer-review

Abstract

We often give advice to influence others, but could our own advice also be shaped by the very individuals we aim to influence (i.e., advisees)? This reverse flow of social influence—from those typically seen as being influenced to those who provide the influence—has been largely neglected, limiting our understanding of the reciprocal nature of human communications. Here, we conducted a series of experiments and applied computational modelling to systematically investigate how advisees’ opinions shape the advice-giving process. In an investment game, participants (n = 346, across four studies) provided advice either independently or after observing advisees’ opinions (Studies 1 & 2), with feedback on their advice (acceptance or rejection) provided by advisees (Studies 3 & 4). Our findings reveal that advisors tend to adjust their advice to align with the advisees’ opinions (we refer to this as the alignment bias) (Study 1). This tendency, which reflects normative conformity, persists even when advisors were directly incentivized to provide accurate advice (Study 2). As feedback is introduced, advisors’ behavior shifts in ways best captured by a reinforcement learning model, suggesting that advisees’ feedback drives adaptations in advice giving that maximize acceptance and minimize rejection (Study 3). This adaptation persisted even when acceptance is rare, as bolstered by the model-based evidence (Study 4). Collectively, our findings highlight advisors’ susceptibility to the consequence of giving advice, which can lead to counterproductive impacts on decision-making processes and misinformation exacerbation in social encounters.

Original language	English
Article number	e1013732
Number of pages	29
Journal	PLoS Computational Biology
Volume	21
Issue number	12
DOIs	https://doi.org/10.1371/journal.pcbi.1013732
Publication status	Published - 2 Dec 2025

Bibliographical note

Author summary
Among the various forms of opinion exchange, advice stands out for its informational richness and its prevalence in word-of-mouth communication. Our research presents a counterintuitive view, suggesting that advice can be considerably biased—particularly by those receiving it (i.e., advisees). Advisors incline to align their opinions (advice) with those of their advisees (we refer to this as the alignment bias), even at the cost of compromising accuracy of their advice. By unraveling the advisors’ reactions to the acceptance/rejection from advisees using computational modeling, our data proposes an evolutionary perspective of how alignment bias emerges: advice-giving behavior can be shaped by advisees’ feedback (i.e., acceptance or rejection of advice). This nuanced bias, while understandable, can lead to poor decisions and spread inaccurate information. Zooming in, this susceptibility to the social outcomes of advice giving potentially leads to counterproductive decision-making and misinformation exacerbation. Zooming out, our work highlights a hidden social dilemma in everyday communication and shows how even well-meaning advice can become distorted by our need to connect with others.

Access to Document

10.1371/journal.pcbi.1013732Licence: Creative Commons: Attribution (CC BY)

LuoX2025AdviseFinal published version, 1.96 MBLicence: Creative Commons: Attribution (CC BY)

1 Preprint

Do we advise as one likes? The alignment bias in social advice giving
Luo, X., Zhang, L. & Pan, Y., 4 May 2024, PsyArXiv.
Research output: Working paper/Preprint › Preprint

Cite this

@article{be9ad3e4cf8b4e9dbbd01094c2803808,

title = "Do we advise as one likes? The alignment bias in social advice giving",

abstract = "We often give advice to influence others, but could our own advice also be shaped by the very individuals we aim to influence (i.e., advisees)? This reverse flow of social influence—from those typically seen as being influenced to those who provide the influence—has been largely neglected, limiting our understanding of the reciprocal nature of human communications. Here, we conducted a series of experiments and applied computational modelling to systematically investigate how advisees{\textquoteright} opinions shape the advice-giving process. In an investment game, participants (n = 346, across four studies) provided advice either independently or after observing advisees{\textquoteright} opinions (Studies 1 \& 2), with feedback on their advice (acceptance or rejection) provided by advisees (Studies 3 \& 4). Our findings reveal that advisors tend to adjust their advice to align with the advisees{\textquoteright} opinions (we refer to this as the alignment bias) (Study 1). This tendency, which reflects normative conformity, persists even when advisors were directly incentivized to provide accurate advice (Study 2). As feedback is introduced, advisors{\textquoteright} behavior shifts in ways best captured by a reinforcement learning model, suggesting that advisees{\textquoteright} feedback drives adaptations in advice giving that maximize acceptance and minimize rejection (Study 3). This adaptation persisted even when acceptance is rare, as bolstered by the model-based evidence (Study 4). Collectively, our findings highlight advisors{\textquoteright} susceptibility to the consequence of giving advice, which can lead to counterproductive impacts on decision-making processes and misinformation exacerbation in social encounters.",

author = "Xitong Luo and Lei Zhang and Yafeng Pan",

note = "Author summary Among the various forms of opinion exchange, advice stands out for its informational richness and its prevalence in word-of-mouth communication. Our research presents a counterintuitive view, suggesting that advice can be considerably biased—particularly by those receiving it (i.e., advisees). Advisors incline to align their opinions (advice) with those of their advisees (we refer to this as the alignment bias), even at the cost of compromising accuracy of their advice. By unraveling the advisors{\textquoteright} reactions to the acceptance/rejection from advisees using computational modeling, our data proposes an evolutionary perspective of how alignment bias emerges: advice-giving behavior can be shaped by advisees{\textquoteright} feedback (i.e., acceptance or rejection of advice). This nuanced bias, while understandable, can lead to poor decisions and spread inaccurate information. Zooming in, this susceptibility to the social outcomes of advice giving potentially leads to counterproductive decision-making and misinformation exacerbation. Zooming out, our work highlights a hidden social dilemma in everyday communication and shows how even well-meaning advice can become distorted by our need to connect with others.",

year = "2025",

month = dec,

day = "2",

doi = "10.1371/journal.pcbi.1013732",

language = "English",

volume = "21",

journal = "PLoS Computational Biology",

issn = "1553-734X",

publisher = "Public Library of Science (PLOS)",

number = "12",

}

TY - JOUR

T1 - Do we advise as one likes? The alignment bias in social advice giving

AU - Luo, Xitong

AU - Zhang, Lei

AU - Pan, Yafeng

N1 - Author summary Among the various forms of opinion exchange, advice stands out for its informational richness and its prevalence in word-of-mouth communication. Our research presents a counterintuitive view, suggesting that advice can be considerably biased—particularly by those receiving it (i.e., advisees). Advisors incline to align their opinions (advice) with those of their advisees (we refer to this as the alignment bias), even at the cost of compromising accuracy of their advice. By unraveling the advisors’ reactions to the acceptance/rejection from advisees using computational modeling, our data proposes an evolutionary perspective of how alignment bias emerges: advice-giving behavior can be shaped by advisees’ feedback (i.e., acceptance or rejection of advice). This nuanced bias, while understandable, can lead to poor decisions and spread inaccurate information. Zooming in, this susceptibility to the social outcomes of advice giving potentially leads to counterproductive decision-making and misinformation exacerbation. Zooming out, our work highlights a hidden social dilemma in everyday communication and shows how even well-meaning advice can become distorted by our need to connect with others.

PY - 2025/12/2

Y1 - 2025/12/2

N2 - We often give advice to influence others, but could our own advice also be shaped by the very individuals we aim to influence (i.e., advisees)? This reverse flow of social influence—from those typically seen as being influenced to those who provide the influence—has been largely neglected, limiting our understanding of the reciprocal nature of human communications. Here, we conducted a series of experiments and applied computational modelling to systematically investigate how advisees’ opinions shape the advice-giving process. In an investment game, participants (n = 346, across four studies) provided advice either independently or after observing advisees’ opinions (Studies 1 & 2), with feedback on their advice (acceptance or rejection) provided by advisees (Studies 3 & 4). Our findings reveal that advisors tend to adjust their advice to align with the advisees’ opinions (we refer to this as the alignment bias) (Study 1). This tendency, which reflects normative conformity, persists even when advisors were directly incentivized to provide accurate advice (Study 2). As feedback is introduced, advisors’ behavior shifts in ways best captured by a reinforcement learning model, suggesting that advisees’ feedback drives adaptations in advice giving that maximize acceptance and minimize rejection (Study 3). This adaptation persisted even when acceptance is rare, as bolstered by the model-based evidence (Study 4). Collectively, our findings highlight advisors’ susceptibility to the consequence of giving advice, which can lead to counterproductive impacts on decision-making processes and misinformation exacerbation in social encounters.

AB - We often give advice to influence others, but could our own advice also be shaped by the very individuals we aim to influence (i.e., advisees)? This reverse flow of social influence—from those typically seen as being influenced to those who provide the influence—has been largely neglected, limiting our understanding of the reciprocal nature of human communications. Here, we conducted a series of experiments and applied computational modelling to systematically investigate how advisees’ opinions shape the advice-giving process. In an investment game, participants (n = 346, across four studies) provided advice either independently or after observing advisees’ opinions (Studies 1 & 2), with feedback on their advice (acceptance or rejection) provided by advisees (Studies 3 & 4). Our findings reveal that advisors tend to adjust their advice to align with the advisees’ opinions (we refer to this as the alignment bias) (Study 1). This tendency, which reflects normative conformity, persists even when advisors were directly incentivized to provide accurate advice (Study 2). As feedback is introduced, advisors’ behavior shifts in ways best captured by a reinforcement learning model, suggesting that advisees’ feedback drives adaptations in advice giving that maximize acceptance and minimize rejection (Study 3). This adaptation persisted even when acceptance is rare, as bolstered by the model-based evidence (Study 4). Collectively, our findings highlight advisors’ susceptibility to the consequence of giving advice, which can lead to counterproductive impacts on decision-making processes and misinformation exacerbation in social encounters.

U2 - 10.1371/journal.pcbi.1013732

DO - 10.1371/journal.pcbi.1013732

M3 - Article

SN - 1553-734X

VL - 21

JO - PLoS Computational Biology

JF - PLoS Computational Biology

IS - 12

M1 - e1013732

ER -

Do we advise as one likes? The alignment bias in social advice giving

Abstract

Bibliographical note

Access to Document

Fingerprint

Research output

Do we advise as one likes? The alignment bias in social advice giving

Cite this