Ask HN: How to figure out if someone is smart during a casual conversation?

Depends on what kind of smarts you are looking for:<p>Quickness? Adaptability or rote responses? Knowledge? Trivia or domain knowledge? Ability to learn new ideas? Ability to analyze? Ability to memorize? Ability to improvise? Etc<p>You could probably ascertain surface level measures for those kind of questions in the course of a single conversation, but you’d only be seeing a sliver of their self in one conversation and it would be best to just ask the person “how smart do you think you are?” in some innocent way at an opportune moment in the conversation.

&quot;Smart&quot; is vague, but in any case I&#x27;m not sure it&#x27;s possible. For sure I can&#x27;t.<p>Here&#x27;s my anecdata with a very narrow but pragmatic definition of &quot;smart&quot;. When I did academic research, there was <i>no correlation</i> between any aspect of the casual conversation I held with students and their abilities in maths and CS theory. N=O(10).

One heuristic not often used is openness to new information and flexibility of opinion. If you can select an area where the person may have a pre-formed opinion that you know is objectively unwarranted with current data, you can present that data and see their reaction. Intelligent people tend to be open to and actively enjoy recieiving new information which challenges and alters their opinion, creating a more informed and more well founded perspective. Insecure, groupthink people will instead vehemently deny the validity of the new information and rigidly stick to a pre-formed opinion. This is a clear sign of a limited capacity for critical thinking.<p>Edit: Just realised Alan Kay said this more succinctly. <i>A change in perspective is worth 80 IQ points.</i>

I start by assuming everyone is smarter than me, then I rule people out by process of elimination when I hear them say dumb things.<p>Also, I think I accidentally trick smart people into thinking I&#x27;m smart sometimes by just saying random things that turn out to have some unexpectedly humorous connection inside the smart person&#x27;s brain. Sometimes I figure out the joke later, but often I don&#x27;t. I think once this has happened people give me the benefit of the doubt when my random shit doesn&#x27;t sync up to their preconceptions and just assume that they didn&#x27;t get the joke this time.

In my experience people don&#x27;t have the ability to determine if someone is smarter than themselves. They can detect that someone is a different intelligence level. Most people will react the same to someone with much greater intelligence as someone with much less intelligence.<p>Secondly, intelligent people modulate their communication to be what the people they are talking to expect. They won&#x27;t use big words at the bar, but will totally change in an academic or professional setting.<p>One way that &quot;smarter&quot; than oneself is revealed is when a person predicts or anticipates something. That could be experience, or just indicates that they had already processed the situation to the same level.

I&#x27;ve been playing around with trying to arbitrarily quantify perceived intellect.<p>Here&#x27;s my working formula at the moment:<p>`P = A * SUM([B * C * TopicA] + [B * C * TopicB] + [B * C * TopicC] [...])`<p>Where:<p>P = Perceived intellect<p>A = Coefficient that measures the ability to connect different concepts<p>B = Depth of knowledge of a topic<p>C = Ability to break down topic into concepts<p>The people I perceive to be smart come up with logically-sound answers to questions they don&#x27;t know the answer to, based on past knowledge.<p>If I ask `what is X?`, they say `I don&#x27;t know what X is, but based on Y and Z, it might be ...`<p>So if you&#x27;re trying to identify if this person can come off as smart, perhaps a tactic could be asking them a question about a domain they are not an expert in.

This rather negative excerpt from Jack Nasher, &quot;Conv!nced - How to Prove Your Competence and Win People Over&quot;, p. 185-187 may be helpful:<p>&quot;We are frightenly bad at making an accurate assessment of other people&#x27;s competence, and the same is true for the closely related factor of intelligence⁰. In one experiment, only 20 percent of people tested were able to assess the intelligence of others with higher accuracy than a random number generator.<p>[…]<p>To assess intelligence, which is colesely related to competence, there are, however, some generally valid factors. The following items are charcteristics of an actual high intelligence¹:<p>• speaking quickly • using easily understandable speech and standard English • making eye contact while talking • displaying self-confident behavior • reacting quickly and with little hesitation<p>[…] Unlike many of the previously discussed points, however, these five items are also indicators of <i>actual</i> high intelligence. All of these points can rather easily be evaluated by mere observation. Interestingly, it is easier for us to make an accurate evaluation if we only hear people and do not see them - visual factors often are misleading.&quot;<p>[0] Reynolds, D.J. &amp; Gifford, R. (2001). The sounds and sights of intelligence: A lens model channel analysis. Personality and Social Psychology Bulletin, 27, 187-200.<p>[1] Murphy, N. (2007). Appearing smart: The impression management of intelligence, person perception accuracy, and behavior in social interaction. Personality and Social Psychology Bulletin, 33, 325-39<p>and<p>[2]

Be interested in and ask interesting questions about the things they are discussing. Someone who is genuinely into a subject will generally be quite happy to talk about it at length.<p>CAUTION: if you do this, you will unmask far more poseurs than genuinely interesting people. Be prepared for the social fallout.

Casually discuss logic puzzles and probe to see if they understand the concept of a paradox. Talk about the odds of something happening and see if they can think in terms of probabilities beyond 50&#x2F;50.<p>I&#x27;ve found myself as a mathematician that conversations just flow with other people who have a math background in ways that they don&#x27;t with other people. It&#x27;s probably due to a conversational deficit on my part.<p>As others have mentioned, something really important is mental flexibility, the ability to look at new evidence and adjust your opinion. You can find out if people have this ability in a casual conversation, but it would have to be a <i>long</i> casual conversation.

I presume everyone is smart. Intelligence is orthogonal to my interests. It&#x27;s orthogonal to my opinions. It&#x27;s orthogonal to my education. It&#x27;s orthogonal to their ability to communicate in a way that is clear to me.<p>And it is orthogonal to all those things in me.

Anyone who listens for more than 20 seconds is over 110 IQ. People who talk mostly about other people rarely are.

Not a good proxy for &quot;smart&quot;, but when I&#x27;m deciding whether or not to pursue future interactions (professional or socially), I turn to the classic turing test. Spend most of the conversation listening, except as needed to keep the conversation going. Imagine you can&#x27;t see or hear them, but only read their words on paper. How sure are you that they are not a computer program? For me, notes of emotion, hints of future hope or past regrets, recognition that humans are flawed or acknowledging that others may hold a different perspective are strong moves. Silliness, reasoning, compassion, specialized interests score points too.<p>Next, run this test on yourself. Be more human.

I&#x27;ve had great conversations with strangers about whether aliens exist, or whether they can be made of sand, or rock, or gas. The intelligent people know how to link that to other things - temperatures, evolution, whether planets or cities can be sentient, the definition of consciousness. You can talk about things like souls or ghosts too.<p>You can also open up fixed topics that everyone has an opinion about. Politics, religion&#x2F;life, economics, and investing are nice. Everyone intelligent has <i>some</i> kind of investment strategy. A lot of people at least try to figure out religion at some point. A lot of intelligence is understanding that you could be wrong, and dealing with that. If someone straight up says, &quot;Communism bad, end of story,&quot; without trying to understand <i>why</i> communism is bad, they&#x27;re likely not very smart.

A telltale sign of being smart in my experience is the habit of waiting until someone else finishes talking before saying anything. I suspect that people who aren&#x27;t smart are conditioned to talk over other people because they&#x27;re accustomed to being justifiably ignored otherwise. The nice thing about this heuristic is that it works just as well for small talk and deep conversations.

What do you mean by &quot;smart&quot;?

Use the One-Joke Mensa Test:<p>&quot;Did you hear the one about the cannibal who passed his friend in the street?&quot;

This isn&#x27;t going to go well for anybody above you, since you are only equipped to judge in one direction. Some ideas though:<p>1. generating new and unusual humor<p>2. spotlessly logical grammar, aside from interruptions like changing an argument mid-sentence<p>3. understands a huge vocabulary (note: doesn&#x27;t necessarily use it)<p>None of those is going to get you very far. It&#x27;s also very verbal-oriented.

when you use somebody as a resource in casual conversation based on your knowledge of their domain expertise and they provide information that you had not considered. That&#x27;s about it

“Smart” is merely one characteristic you can scan for in a conversation. Find at least a pair: smart &amp; sociopathic, smart &amp; empathetic, smart &amp; withdrawn, smart &amp; searching, smart &amp; engaged, etc.

Put a piece of paper trash on the floor and see if they put it in the waste basket upon entering the room. Tests values.

ask them to expand on any of their views&#x2F;probe for more nuanced opinions

Ask him which operating system he uses.

This is a simple supervised learning classification problem, just select the right features. You have to know what kind of smartness you are looking for, for your training data set. Record the conversation then pull it through your classifier.