The Library

This is one of the studies that started a whole field. Participants made an estimate, saw an adviser’s estimate, and then gave a final answer. On average, people moved towards the advice, but not very far.

The authors’ explanation is simple. We can see all the reasons behind our own view, and almost none behind someone else’s. So our own view feels heavier.

They also found that advisers’ reputations form quickly, and that a good reputation is easier to lose than to gain. A later review of two decades of this research (Bonaccio and Dalal, 2006) found the same tendency to underweight advice across many kinds of task.

When two people disagree about a number, the most accurate answer is usually somewhere between them. Soll and Larrick found that most people do not act on this. They tend to pick one estimate, usually their own, and discard the other.

Averaging works because two people’s errors often point in different directions, and partly cancel out. Choosing one view keeps all of that person’s error.

The practical lesson is gentle: when someone you respect disagrees with you, it is worth asking whether the answer lies between you before deciding who is right.

We tend to trust people who sound sure of themselves. Earlier research called this the confidence heuristic. This study asked what happens when people can also see whether the adviser was right.

When accuracy was visible, confident advisers who were wrong lost credibility. When it was hidden, or cost something to find out, confidence won, whether or not it was deserved.

There was a further twist. The more confident the adviser sounded, the less interested people were in paying to check their track record.

The researchers kept the words identical and changed only the label. People who believed a human had written the reply felt more understood and more supported.

Some participants were willing to wait for a human reply rather than receive an AI reply straight away.

We read this as a finding about people more than about technology. Part of what makes advice land is the belief that another person took the time to think about us.

The study had two parts. First, the researchers compared how AI models and people responded to the same personal dilemmas. The models sided with the user far more often.

Second, they let real people discuss conflicts, including their own, with either a flattering model or a more balanced one. After a single conversation with the flattering model, people felt more certain they were right and less willing to repair things with the other person.

They also rated the flattering model as more trustworthy, and were more likely to come back to it. The same pull exists in human advisers, which is why this study sits next to the older texts on flattery.

Disclosure is the usual answer to conflicts of interest: tell the client how you are paid, and let them adjust. This study found that it can work less well than we assume.

Clients did not discount biased advice enough, even after being told about the bias. And some advisers, having disclosed, gave more biased advice, as if disclosure had settled their obligation.

Later work by some of the same researchers found the effect depends on context, so it is fair to say disclosure can backfire, rather than that it always does. Knowing how an adviser is paid still matters. It is a starting point for questions, and the questions do the real work.

Audit studies are borrowed from research on discrimination. You send carefully prepared people into real offices and record what happens.

Here the shoppers posed as ordinary investors. Some held low-cost index funds, some chased last year’s winners. The advisers rarely corrected the costly habits, and often encouraged moves that generated fees.

It is a working paper rather than a journal article, and one city is not every market. It is still one of the clearest looks we have at what happens inside the room.

The researchers could see both the clients and the portfolios their advisers built. They expected portfolios to vary with each client’s age, goals and appetite for risk.

They found that the adviser mattered more than the client. Each adviser tended to put clients into similar portfolios, whoever they were.

Advice did get people investing in shares, which has real value. But fees took away the whole of that benefit for the average client. The study is a reminder to ask how advice is tailored to you, and what it costs each year.

In the United States, complaints and disciplinary actions against advisers are publicly recorded. The researchers assembled this record for the whole industry.

Most advisers had a clean record. But misconduct was concentrated, repeated, and often followed the adviser from firm to firm. It was also more common in areas with older, wealthier or less educated populations.

Many countries keep similar public registers. Looking an adviser up takes a few minutes, and it is a perfectly polite thing to do.

The team compared the diagnosis each patient arrived with against the diagnosis Mayo reached. In about 12% of cases the original diagnosis was confirmed as it stood. In about two-thirds it was refined or redefined. In 21% it changed completely.

These were complex cases that had already been referred on, at one centre, so the numbers do not describe ordinary medicine. Some headlines turned the study into “88% of diagnoses are wrong”, which goes further than the data.

What it does show is that a second look by a fresh pair of eyes often changes the picture, especially when a problem is complicated.

Many of us put off asking for advice because we expect to be turned down, or to be a burden. This research suggests the fear is exaggerated.

Participants consistently expected more refusals than they got. The authors’ explanation is that askers think about the cost of saying yes, and forget how awkward it is to say no to someone’s face.

If there is someone whose view you would value, the odds that they will make time for you are better than you think.

We usually assume the person receiving advice is the one who benefits. This study looked at the giver.

People who were struggling with a goal felt more motivated after giving advice than after receiving it, even though most people predicted the opposite. Advising someone else reminded them of what they already knew, and made them more confident they could act on it.

A later field experiment by Eskreis-Winkler and colleagues (PNAS, 2019), with 1,982 high-school students, found a modest improvement in grades among those who gave advice to younger students. Advice runs in both directions.

This is not a study about advice as such. It is about the web of relationships through which most advice reaches us: family, friends, colleagues and mentors.

Pooling data from many countries, the authors found that the strength of people’s social ties was linked to how long they lived, by a margin comparable to well-known health risks.

The studies were observational, so they show a strong association rather than proof of cause. Still, it is a sobering reminder that the people we can turn to are worth tending.

Given the choice, many of us send a message rather than call. These experiments suggest we underestimate what we lose.

Hearing someone’s voice made people feel more connected, and the awkwardness they feared mostly did not happen.

For advice in particular, this is worth remembering. Tone carries a great deal: hesitation, warmth, conviction. A call often says more than a careful paragraph.

Strangers were asked to discuss either small-talk topics or deeper questions, the kind that come up when someone asks for advice. Before talking, people expected the deep conversations to be awkward.

Afterwards they reported the opposite. The conversations felt less awkward and more connecting than predicted, and people liked each other more.

This matters for advice, because the most useful conversations are often the ones we avoid starting.