{"id":925,"date":"2021-07-01T16:00:00","date_gmt":"2021-07-01T16:00:00","guid":{"rendered":"https:\/\/resources.illc.uva.nl\/illc-blog\/?p=925"},"modified":"2021-07-02T08:26:23","modified_gmt":"2021-07-02T08:26:23","slug":"inside-a-reasoners-head","status":"publish","type":"post","link":"https:\/\/resources.illc.uva.nl\/illc-blog\/inside-a-reasoners-head\/","title":{"rendered":"Inside a reasoner’s head"},"content":{"rendered":"\n

02 July 2021, Iris Proff<\/a><\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

\u00a9 Iris Proff<\/p>\n\n\n\n

Logic is what makes your computer work and it is the scaffold of any scientific discipline. Clearly, logic matters \u2013 but does formal logic describe how we solve problems in our daily lives? Are we evaluating logical formulas in our heads when we decide who to vote for, which job to apply for, or which route to take to the supermarket?<\/p>\n\n\n\n

There are many hints that we don\u2019t. Human reasoning is imperfect, messy, inconsistent. Therefore, logicians often consider empirical findings irrelevant for their studies. However, the reasoning mistakes humans make are not arbitrary, but very systematic. Slowly, logicians are starting to recognize that the realm of human reasoning can teach us many things about logic \u2013 and vice versa.<\/p>\n\n\n\n

Anthi Solaki<\/a> is one of them; she recently completed her PhD at the ILLC<\/a>, during which she bridged the distinct worlds of formal logic and psychology of reasoning. \u201cI am making logical models more realistic by taking into account human limitations,\u201d she says.<\/p>\n\n\n\n

But where exactly do those limitations lie?<\/p>\n\n\n\n

The mother of reasoning tasks<\/strong><\/h2>\n\n\n\n

Whoever took a course in basic logic can confirm: filling in truth tables, juggling with logical formulas and drawing Venn diagrams are skills we can learn \u2013 but they do not intuitively match our everyday-reasoning. In real life, people fall for all kinds of logical fallacies<\/a>. Someone who believes to be \u201con a run\u201d when playing roulette falls prey to the \u201cgamblers fallacy\u201d. Election campaigns are often the scene of \u201cad hominem fallacies\u201d where one side attacks a personal feature of their opponent instead of their argument. And when you slightly misinterpret what your opponent said to have a better ground for your attack, you are making use of the \u201cstrawman fallacy\u201d.<\/p>\n\n\n\n

People even struggle to implement basic logical rules. This was famously demonstrated by the Wason Selection Task, first developed by Peter Cathcart Wason<\/a> in the 1960s.<\/p>\n\n\n\n

\"\"<\/figure>
\n

The Wason Selection Task<\/strong><\/p>\n\n\n\n

Consider these four cards. Each card has a number on one side and a letter on the other. Which cards do you have to turn around to check whether the following rule holds: if a card has a vowel on one side, then it has an even number on the other?<\/em><\/p>\n<\/div><\/div>\n\n\n\n

To get the puzzle right, a reasoner must apply two logical rules. She needs to check the card with the vowel to confirm that there is an even number on the other side, which is intuitive to most. But she also needs to check the card with the uneven number to confirm that there is no vowel on the other side, which proves to be a harder reasoning step. 90% of people get the answer wrong.<\/p>\n\n\n\n

Interestingly, later experiments showed that when the same task is phrased within a familiar context, people perform much better. This is especially true when they need to reason about who might be violating a social rule<\/a>. These findings gave rise to a heated debate in the field: How do humans reason? And why are some forms of reasoning harder than others?<\/p>\n\n\n\n

\"\"
Anthi Solaki will further pursue her research as a postdoctoral researcher at the ILLC within the research priority area Humane AI<\/a> of the University of Amsterdam.<\/figcaption><\/figure><\/div>\n\n\n\n

Inside a reasoner\u2019s head<\/strong><\/h2>\n\n\n\n

\u201cThere are different schools of thought in psychology about how people reason,\u201d says Anthi Solaki. Some psychologists argue that people represent content-independent, logical rules in their minds which they apply when solving a reasoning problem. Others argue that people only learn statistical patterns from experience. Think about how you can probably not spell out all the grammatical rules of your native language, while being perfectly able to apply those rules in practice. Again others claim that our reasoning ability evolved specifically in contexts in which it is adaptive \u2013 such as detecting that someone is breaking a social rule.<\/p>\n\n\n\n

Finally, dual process theories propose that humans can do both: They can solve problems automatically using patterns and biases they learned from experience. And they can solve problems by applying abstract rules \u2013 but this process is effortful; it costs time and working memory. Depending on how hard a problem seems, how much time is available and how much is at stake, a reasoner can choose between these two reasoning modes.<\/p>\n\n\n\n

Should logic care about psychology?<\/strong><\/h2>\n\n\n\n

The question if humans are able to reason according to abstract rules is at the heart of the so-called Rationality Debate. Kant, Bolzano and Frege argued that human reasoning with all its biases and flaws cannot tell us anything about the correct<\/em> way to reason. According to their \u201canti-psychologistic\u201d stance, the study of logic should be not bothered with empirical findings. This dogmatic conviction created a strict separation between research in logic and psychology. But there is a case to be made against this separation. In a paper from 2008<\/a>, the logician and co-founder of our institute Johan van Benthem poetically raises the question:<\/p>\n\n\n\n

\u201cDo the empirical facts about human reasoning matter to logic, or should we just study relationships between proof patterns, and the armies of them that we call formal systems, in some eternal realm where the sun of Pure Reason never sets?\u201d<\/strong><\/p>\n\n\n\n

Logic and psychology of reasoning certainly serve different purposes, van Benthem says. Logic aims to develop normative rules \u2013 how we should<\/em> reason \u2013 while psychology describes how we do in fact<\/em> reason. Still, van Benthem argues that logic can inspire interesting psychological experiments and that findings from psychology can inspire research in formal logic. He concludes: \u201cIf logical theory were totally disjoint from actual reasoning, it would be no use at all, for whatever purpose!\u201d<\/p>\n\n\n\n

Epistemic logic and the imperfections of human reasoners<\/strong><\/h2>\n\n\n\n

Epistemic logic formally describes the knowledge and belief states of agents and how these can evolve. It certainly is concerned with the reasoning of humans, rather than remaining in what van Benthem called the \u201ceternal realm of Pure Reason\u201d. Still, models of epistemic logic make assumptions that real human agents cannot live up to, argues Anthi Solaki:<\/p>\n\n\n\n