In the 18th century, philosopher James Beattie compiled a list of 17 common-sense beliefs. A few are incontrovertible: âI existâ; âA whole is greater than a partâ; âVirtue and vice are differentâ. But others seem unnecessarily moralising: âIngratitude ought to be blamed and punishedâ; âI have a soul distinct from my bodyâ; âThere is a Godâ. Then, there are the scientifically contestable: âThe senses can be believedâ; âI am the same being that I was yesterday â or even 20 years agoâ; âTruth existsâ. Overall, his list seems quaint and outdated. Worse still, it gives no clear idea of what common sense is. Surely, we can do better.
Superficially, common sense seems easy to define: it is generally seen as knowledge or beliefs that are obvious â or should be obvious â to everyone. Yet it is strangely difficult to pin down. Often portrayed as universal, it is also often claimed not to exist. With that in mind, it might surprise you to hear that nobody has tried to measure the âcommonnessâ of this knowledge or its intrinsic properties (its âsensicalityâ) â until now. Shockingly, this research shows that common sense may not be common at all.
Advertisement
If true, the implications are huge. From parenting to politics and from public health to law, what counts as common sense matters. Increasingly, it is also a technological issue, with computer scientists keen to instil it in artificial intelligence-driven robots to make them behave more like people. So, what do we know about this slippery concept? And how can new insights help us to understand each other better, or build machines we can relate to more easily?
Appeals to common sense . In 1776, they kicked off the US Declaration of Independence: âWe hold these truths to be self-evident, that all men are created equal, that they are endowed by their Creator with certain unalienable RightsâŚâ. Today, judges and doctors regularly rely on common sense when making decisions or issuing advice that could save â or blight â peopleâs lives. Politicians, too, are forever banging on about âcommon-sense approachesâ to national problems. The last UK government even had a minister for common sense. And common-sense reasoning is what we frequently urge children to use â or scold them for lacking. Even so, there is remarkably little research into this everyday concept.
Studies have tended to fall into one of two types. There are those by philosophers and sociologists that , of common sense to people and society. And, more recently, researchers have begun to discuss the desirability and feasibility of . âWhat these otherwise very different genres share is that they both take for granted that we know what common sense is, and that it is indeed common,â says at the University of Pennsylvania. So, he and his colleague set out .
Evaluating common sense
First, they gathered together 4407 statements that someone, somewhere has claimed to be either common sense or self-evidently true. These came from a range of sources: two yearsâ worth of Google News articles, US political campaign emails, popular aphorisms, two collections of material used to train AI called and , and two datasets that the researchers generated themselves â one by getting online participants to write common-sense claims in 13 categories of life, the other by giving people a half statement and asking them to complete it in a common-sense way. The final list ran the gamut from âAll people are humanâ and âGuns are dangerousâ to âYou are what you eatâ and âIf Alex hits Steven, Steven will be upsetâ.
A total of 2046 people were each shown a subset of 50 statements and asked if they agreed with them, and also whether they thought most other people would. Their answers varied, as you might expect given that people donât appear to possess equal quantities of common sense. Watts and Whiting deemed raters whose opinions were those of the majority and who also showed good awareness of the majority opinion to have more common sense than others.
Surprisingly, they found that demographic factors, including age, level of education and political leaning, werenât linked to an individualâs stock of common sense. And there was only a small association with their scores on an intelligence test. âHaving potentially more information about âthe truthâ, due to being more educated or a more careful thinker, for example, seems not to consistently improve peopleâs ability to understand what other people think about issues,â says Watts.
But the study did reveal one thing shared by people with abundant common sense. The researchers assessed the social perceptiveness of participants using a test called . This entails examining a series of photos, each zoomed in on a personâs eye region, along with a short list of emotions and attitudes, and trying to pick the one that describes what that person is feeling. People with more common sense were particularly good at this. âWe suspect this indicates that perceiving what other individuals think is predictive of perceiving what society thinks about an issue,â says Whiting.
The elusive nature of common sense raises problems for instilling it in AI
Intriguingly, these findings hint at what kind of intelligence common sense might be. at Western University in Canada sees parallels with so-called implicit intelligence. Unlike explicit learning, which refers to knowledge that we acquire deliberately, implicit learning entails the unconscious acquisition of abstract knowledge. For example, if you were to ask a 5-year-old â or even some adults â to explain the grammatical rules of their language, they would probably have no clue where to start, yet they do know them well enough to form meaningful sentences because they have acquired them via implicit learning. Kalra and her colleagues have found that, like common sense, , and it is also .
So, someone may have a high IQ but be poor at unconsciously acquiring knowledge about the world, while for another person the opposite is true. Unfortunately, there is no research investigating peopleâs implicit learning ability and their skill at Reading the Mind in the Eyes. However, Kalra thinks it is entirely possible that they are linked, because people who are better at this test may be more influenced by other peopleâs emotional states, which could lead them to agree more with others and so have more common sense. âBut this is speculative,â she adds.
Evidence that some people have more common sense than others isnât particularly surprising. But a second finding made by Watts and Whiting is. They reasoned that, for a given statement to count as common sense, the majority of people would have to agree that it was, and individuals would also need to be aware that other people viewed it as such. Very few claims made the cut. And those that did werenât the type enshrined in the US Declaration of Independence. Instead, they tended to be plainly worded, fact-like statements about objective reality, such as âPlants need nutrients in the soil for them to growâ and âPets should not be taken out during excessive heatâ.
In general, a small fraction of people agreed on a small fraction of statements, leading the pair to describe common sense, as it is usually understood, as âsomewhat illusoryâ. In their paper, they conclude: âCommon sense attains its legitimacy through the perception of its commonness, yet judgements about what is and is not common sense are in practice idiosyncratic and subjective.â They had anticipated some disagreement between people. âBut the extent to which common sense is uncommon is still very surprising,â says Watts.
This is curious, given that most people take the existence of common sense for granted. However, it might be explained by one of humanityâs many psychological biases, according to psychologist at the University of Georgia, who also doubts the commonness of common sense. The describes the finding that people tend to believe they see the world objectively, and therefore their views must not only be right but should also be held by all sensible people. Indeed, Watts and Whiting found that raters tended to believe their views matched those of others, regardless of the actual level of consensus. Another recent study even showed that in the face of clear evidence that someone has listened carefully to our opinions, if they then disagree with us, most of us are convinced .
Idiosyncratic behaviour
If common sense is idiosyncratic, this has big implications outside the lab in all sorts of situations that call for it. As an example, vanDellen highlights disputes during the covid-19 pandemic. Citizens of many countries were urged to use common sense to guide their behaviour. But, she notes, for some people it was common sense to go on holiday with friends who were being careful, whereas for others it was common sense not to interact with anyone outside the home.
From a public health perspective, the advice to âuse your common senseâ is dangerous in two ways, she argues. Firstly, it doesnât tell people exactly how to behave. Secondly, it can breed animosity. âIf I think we are all acting in accordance with common sense, but you are acting differently than me, it can create a sense of hostility and conflict,â she says.
Besides causing friction in human interactions, the elusive nature of common sense also raises all sorts of problems for anyone wanting to instil it in machines. However, Watts and Whiting see a ray of hope here. They believe that although the store of shared common-sense knowledge might be slim for large groups, such as a national population, this might not be the case for certain subsets of these groups. For example, judges may share a pool of legalistic common sense, and likewise for doctors or other medical professionals when it comes to medical know-how. If that is the case, training specialist legal or medical AI bots to have the necessary common sense would be more feasible.
However, for , a computer scientist at the University of Toronto, when it comes to training AI there is no hard line between topics that are common and those that represent expert knowledge. âAnd I donât think there is much to be gained â in computer science, anyway â by trying to draw any such line,â he adds.
Instead, Levesque argues for a different approach in his book â written with fellow AI expert at Cornell University in New York. âWe think of common sense as the ability to make effective use of ordinary, everyday knowledge in achieving ordinary, everyday goals,â he says. For example, if you are driving a familiar route to the grocery store and come across a traffic light that happens to be stuck on red, you will quickly realise the problem and find a workaround. By contrast, an autonomous vehicle lacking common sense will just sit there, waiting. With this pragmatic definition, Levesque and Brachman argue, common sense could be built into machines in the ânot-too-distant futureâ.
at the Massachusetts Institute of Technology is one researcher working on this right now. Most robots are programmed to do one or a small number of fixed things, so they canât adapt to new scenarios, he says. Housework robots, for example, are usually programmed to copy the motions of a human trainer, and they do this very well â until small errors accumulate, or something unexpected happens. Wang and his colleagues aim to get around this by with knowledge contained in AIs called large language models about the subtasks within a given task â cleaning a kitchen, say. This allows the machine to recognise which stage it is on. Then, if it encounters a problem, it can make a new plan and recover on its own. Using LLMs to give robots some degree of physical, practical common sense is a hot topic in the research world, says Wang, âand progress will come soonâ.
Meanwhile, Watts and Whiting are extending their research in an online project (see How much common sense have you got?, below). They hope that clarifying the distinction between what is, and is not, common sense could help to ease tensions between people with opposing views on hot-button issues. âIf I think itâs âjust common senseâ that guns should be regulated and you think itâs common sense that they should not be, the only recourse either of us has is to think there is something wrong with the other person,â says Watts. A clearer delineation between what is common sense and what is personal, subjective opinion could make these disputes more rational, he hopes.
Whatever this ongoing research reveals, the notion of common sense looks set to stay. âItâs a human tendency to believe that our ideas are shared with others and unaffected by our experiences and culture,â says vanDellen. âPeople really resonate with the phrase âcommon senseâ.â
How much common sense have you got?
"Happiness can spark a desire to sing." Do you agree? And do you think most people would feel the same way? What about, "Nobody wants to be in a traffic jam" or "Curiosity makes you want to learn more"? These statements are among many featured in a major ongoing study to explore what constitutes common sense â and whether it even exists. Anyone can participate online at commonsense.seas.upenn.edu
The researchers behind the study, Duncan Watts and Mark Whiting at the University of Pennsylvania, recently found that common sense is surprisingly uncommon (see main story). This new project will allow them to explore the subject further. They hope to understand, on a global level, the types of knowledge that can be considered common sense â and how language and culture influence what knowledge people consider to be self-evident.
