There is a quiet bargain people often make in a conversation with someone new: I will not ask anything that might expose you, as long as you do not ask anything that might expose me.
The result is safe. It can also leave two interested people discussing traffic, job titles and how they know the host while each privately waits for the conversation to become real.
A paper led by Michael Kardas tried to measure the prediction beneath that restraint. Across a series of experiments, people expected personal conversations with strangers to be more awkward, less enjoyable and less connecting than those conversations actually became.
This is one research programme, not settled consensus. Its deeper finding is narrower than “small talk is bad”. People seemed to understand their own interest in another person’s inner life while underestimating how much interest the other person would return.
What the researchers asked strangers to discuss
The paper, “Overly Shallow?”, was published online in 2021 and appeared in the 2022 volume of the Journal of Personality and Social Psychology. Kardas, Amit Kumar and Nicholas Epley reported 12 experiments involving more than 1,800 participants.
Most conversations paired strangers. Some participants received shallow questions about television, the weather or a bad haircut. Others received questions such as what made them feel most grateful, what they would change about the way they were raised, or when they had last cried in front of someone.
Before talking, participants predicted how awkward the exchange would be, how connected they would feel and how much they would enjoy it. Afterward, they rated the experience itself.
The headline compresses several designs into one sentence. Not every experiment simply instructed people to “skip the small talk”. Some compared separate shallow and deep groups. Some had the same person hold one conversation of each type with different partners. One asked participants to generate their own shallow and deep questions rather than using prompts written by the researchers.
That variation matters because the central mismatch survived more than one artificial question list. Even when people created the topics themselves, their forecasts about personal conversation were too pessimistic.
The forecasting error was larger around depth
Both shallow and deep conversations generally went better than people expected. They felt less awkward and produced more enjoyment and connection than participants had forecast.
The error was especially pronounced for deeper conversations. Participants overestimated their awkwardness more than they did for shallow exchanges. When people experienced both types, they felt more connected to the deep-conversation partner and preferred that exchange more than they had expected beforehand.
This does not mean the personal conversations were free of awkwardness. In some comparisons they were slightly more awkward than shallow ones in actual experience. Nor did every deep-versus-shallow comparison produce a statistically clear difference in connection.
The more consistent result was miscalibration. People anticipated a larger social cost and a smaller reward than they subsequently reported. The barrier existed partly in the forecast, before either stranger had given the other a reason to retreat.
I recognise that forecast from work dinners and events. A more interesting question occurs to me, then I edit it into something safer because I assume the person opposite will experience my curiosity as intrusion. The study does not prove that my original question would always land well. It suggests my confidence about the other person’s disinterest may be doing more work than their actual response.
Expected indifference appears to be part of the mechanism
The researchers repeatedly asked participants how caring and interested they expected their partner to be. People tended to underestimate both.
That gap helped explain why their emotional forecasts were wrong. Participants seemed to imagine that a meaningful disclosure might be met with less attention than it actually received. When the conversation partner listened with interest, the anticipated awkwardness had less room to develop.
The mismatch was smaller when people imagined speaking with close friends, relatives or partners. Care is easier to predict when a relationship has already supplied years of evidence. A stranger gives us less information, and uncertainty is easily filled with a cautious assumption.
Two later experiments tested whether those expectations influenced topic choice. Participants encouraged to expect a caring partner selected deeper questions than those led to expect less care. Giving people information about the general tendency to underestimate strangers’ interest also shifted their choices towards more personal topics.
That does not establish perceived care as the only cause. The measures were self-reports collected within structured experiments, and changing a prompt is not the same as changing a person’s durable beliefs. It does show that expectations were connected to what participants were willing to ask.
A related error appears after ordinary first conversations
The pattern sits beside a separate finding known as the liking gap.
In a 2018 Psychological Science paper, Erica Boothby and colleagues found that people often underestimated how much a new conversation partner liked them. The effect appeared after laboratory conversations, at a personal-development workshop and among first-year university suite mates getting to know one another over several months.
I have written about that study before. What connects it to the deeper-conversation experiments is the direction of the mistake. We have direct access to our own hesitations, but only partial access to another person’s warmth. Their composure can look like indifference while our own composure hides the same uncertainty from them.
Neither paper shows that strangers secretly want unlimited intimacy. Together, they describe a more modest social blind spot. People can be systematically conservative when estimating how positively another person will receive them.
Small talk still provides useful information
The obvious overreading would be to treat shallow conversation as a failure and personal disclosure as the correction. The evidence does not justify that.
Small talk lets people test pace, reciprocity and attention without much risk. A question about the weekend gives the listener room to answer briefly or open a door. Consent was also built into the Kardas experiments: participants knew they had agreed to discuss a set of questions. Asking a stranger in a supermarket queue when they last cried is a different social event.
Naturalistic evidence also complicates the case against small talk. A 2018 pooled analysis of 486 people found that more substantive daily conversation was moderately associated with greater life satisfaction, but it found no reliable association between small talk and life satisfaction. The study was observational, so it could not establish that conversation type caused wellbeing. It did show why “less small talk” should not be smuggled into the findings as a general rule.
The Kardas paper measured brief, structured encounters, mostly in the United States, using participants’ own ratings immediately before and after. It did not test whether rapid disclosure produces lasting friendship, whether the effect travels unchanged across cultures, or whether every personal topic is appropriate.
A useful response is smaller than forced intimacy. One can ask a question with somewhere real to go, then notice whether the other person goes there. “What do you do?” can become “What part of it keeps your attention?” A question about where someone lives can become what they have come to value about living there.
The other person may close the door, and that answer deserves respect. The research simply gives less reason to decide for them before asking.
