The Same Number, a Different Source: Advice with an Expert Label

A rental estimate may seem more persuasive when an experienced estate agent supplies it than when an inexperienced person offers the same number. In ordinary life there can be a sensible reason: expertise may actually provide better information. But what happens if the information’s quality stays constant while only the description of its source changes? That question does not require deciding in advance whether trusting experts is virtuous or mistaken. It allows the ingredients of an advice-taking decision to be separated.

Dar Meshi, Guido Biele, Christoph W. Korn and Hauke R. Heekeren examined this distinction in a rental-estimation task published in 2012. Its defining feature was that every piece of advice presented as coming from either an expert or a novice was actually the listed rental price of the relevant apartment. The two conditions therefore did not differ because one supplied better numbers. The researchers considered the value assigned to advice, the gap between the advice and an initial opinion, and the amount by which the final response changed. [1]

An opinion existed before the advice arrived

The study recruited 29 native German speakers aged twenty to thirty. None had previously rented an apartment in New York City or worked in the property industry. That selection gave the task a particular meaning. Participants could form estimates, but did not have the experience attributed to the local experts introduced to them. This was not a professional examination of estate agents. It was a controlled task in which people dealt with uncertainty and a described source of outside information.

Materials came from real New York rental listings collected in September 2010. Participants saw four attributes: floor area in square metres, number of rooms, number of bathrooms and a three-level neighbourhood index. Prices were converted to euros. Before the main task, one training session introduced example prices, while another familiarized participants with the procedure. Training apartments were not reused in the main experiment. This preparation supplied some basis for estimation without giving participants the market experience attributed to the experts.

Each participant encountered 120 main-task trials. First came the apartment’s attributes and an initial rental estimate. Next, a display announced whether an expert, a novice or no advisor would be involved. After a delay, the advice amount appeared alongside the participant’s first estimate, and a final estimate was requested. In the no-advice condition, participants reconsidered their answer without receiving a new number. The behaviour of interest was therefore a revision of an existing judgment, not the first selection of a number by someone who had yet to form an opinion.

The source’s description was separate from its accuracy

Participants were told that expert advice came from New York estate agents, whereas novice advice came from people with experience similar to their own. Both kinds of advisor were described as being paid for giving estimates close to the actual rental price. The researchers used this account to reduce the possibility that one source would be considered deliberately unhelpful, while making expertise the prominent distinction. That procedure does not prove that expertise and trustworthiness are always separable in life; it explains how the experiment attempted to distinguish them.

Every supplied recommendation was the relevant listing’s price. “Actual price” in this task meant that advertised amount, not an independently documented rent agreed after negotiation. Participants received no feedback about their accuracy during the main experiment. They therefore could not use successive outcomes to learn which source had been reliable. Accuracy bonuses encouraged effort, but the available description of the advisor remained expert or novice. Participants were debriefed about the experimental arrangement afterwards. [1]

This is both an advantage and a boundary of the design. Keeping recommendation quality constant helps expose the influence of the expertise label. Removing ongoing accuracy feedback, however, does not represent a long relationship in which someone observes an advisor’s successes, mistakes or changing reliability. Source evaluation here depended on the reputation supplied to participants, rather than a history of outcomes they could inspect for themselves.

Influence was measured relative to the initial gap

The researchers calculated the distance between the initial estimate and the recommendation, then expressed revision as a weight of advice. To illustrate the logic independently, suppose the initial estimate is €1,800, the recommendation is €2,200 and the final estimate is €2,000. The person has moved halfway across the €400 gap. This is an explanatory example, not a reported participant’s response. Keeping the initial answer and adopting the recommendation completely correspond to different positions on the measure.

Moving away from advice was not treated as equivalent to moving towards it. Negative weights were retained so that the direction of revision remained visible. Conversely, the analysis excluded unanswered trials, unusual response times, answers with no revision and movements beyond a specified upper limit. The removal of zero-change responses is especially important to interpretation. These results should not be treated as an exhaustive measurement of the probability that a person accepts any offered recommendation. The analysis was focused on the degree of influence when advice was used, rather than fully combining that process with the decision not to change an answer.

Behaviourally, participants used recommendations labelled expert more than recommendations labelled novice. Afterwards they also rated the expert advice as more valuable. Weight of advice was higher when the recommendation was close to the initial estimate than when it was distant. Closeness here is not the same as participants learning that they were right. It describes disagreement with their own starting position. Low and high disagreement were also not defined by a universal euro threshold: differences were ranked within each participant’s advice condition and divided into two groups.

Neural associations were not certificates of truth

Functional magnetic resonance data were collected during the task. When participants learned that expert advice would be forthcoming, before they saw the actual recommended number, signal changes in the ventral striatum were greater than in the novice condition. The timing is consistent with a source-related expectation being established before the concrete recommendation arrives. During advice use, average signal changes in the left lateral orbitofrontal cortex were also associated with participants’ average weight of advice. [1]

It would be an unnecessary leap to call this a single brain centre for believing experts. The experiment did not alter a brain region to establish that it was a necessary cause of the behavioural effect. It related task conditions, signal changes and responses. Nor does finding that advice is valued establish that its content is objectively more accurate. In this study, recommendations under both source descriptions carried the same kind of correct target information. The neural observations enriched the account of when the task’s processes occurred; they did not certify the quality of the advice.

The study’s evidence supports treating advice-taking as more than a single event of listening or refusing to listen. An expectation about the source can precede the information, the information can then be compared with an existing opinion, and the opinion can be revised to a particular degree. A narrow estimation task with 29 participants does not establish a universal rule for every culture and decision. It nevertheless shows concretely why supplying an accurate number may not fully explain that number’s influence. Keeping the description of a source separate from the grounds for its recommendation is a cautious thinking principle one can draw from this experiment, not a tested guarantee for every real-world choice.

Sources

[1] Meshi, D., Biele, G., Korn, C. W., and Heekeren, H. R. (2012). How Expert Advice Influences Decision Making. PLOS ONE, 7(11), e49748. Primary research; participants, procedure, advice information, trial exclusions, behavioural findings and selected neural results examined directly. https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0049748

The source is published under a Creative Commons Attribution licence. This article is independently written, not a copy or translation of the source prose. The numerical illustration was constructed for explanation. Source access and factual review: October 14, 2026.