RQ, the rationality quotient
The RQ, the rationality quotient (MIT Press)
Keith E. Stanovich & al. (toward a test of rational thinking) – 2015
Why are we surprised when smart people act foolishly? Smart people do foolish things all the time. Misjudgments and bad decisions by highly educated bankers, for example, brought us the financial crisis of 2008. Smart people do foolish things because intelligence is not the same as the capacity for rational thinking.
The rationality quotient (RQ) explains that these 2 traits, often and incorrectly thought as one, refer to different cognitive functions. The standard IQ test doesn’t measure any of the broad components of rationality (adaptive responding, good judgment and good decision making).
Rational thinking, like intelligence, is a measurable cognitive competence (with the RQ).
Rationality is a more encompassing concept than intelligence
The theory of intelligence (Cattell 1963, 1998) presents 2 dominant factors. Fluid intelligence (Gf) reflects reasoning abilities operating across of variety of domains – including novel one. It is measured by tests of abstract thinking such as figural analogies. Crystallized intelligence (Gc) reflects declarative knowledge measures. It is measured by vocabulary tasks, verbal comprehension, and general knowledge measures. 2 aspects of intelligence are intelligence as process and intelligence as knowledge. However, this model, as numerous, ignore a critical level of cognitive analysis that is important for rationality.
First step: the dual-process theory.
System 1 / System 2 (Kahneman 2011)
The defining feature of Type 1 processing is its autonomy. Type 1 processes execute automatically upon encountering their triggering stimuli. This execution does not depend on input from high-level control systems. Type 1 processing tends to be rapid and associative. It includes emotional regulation, encapsulated modules for solving specific adaptive problems. Type 1 processing also encompasses unconscious implicit learning and conditioning.
In contrast with type 1 processing, Type 2 processing is relatively slow and computationally expensive. One of the most critical functions of type 2 processing is to override Type 1 processing. This is sometimes necessary because autonomous processing has heuristic qualities.
Heuristics and biases
The term “biases” refers to the systematic errors that people make in choosing actions and in estimating probabilities, and then the term “heuristic” refers to why people often makes these errors – because they use mental shortcuts (heuristics) to solve many problems.
Type 1 processing heuristics depends on benign environments to provide obvious cues that elicit adaptive behaviors. In hostile environments, reliance on heuristics can be costly.
Intelligence tests do not focus on the autonomous Type 1 processing of the brain. However, this processing is the most used in our everyday life.
Once detection of a conflict between the normative response and the response triggered by system 1 has taken place, Type 2 processing must display at least related capabilities in order to override Type 1 processing. One is the capability of interrupting Type 1 processing (inhibitory mechanisms). But suppressing one response is not helpful unless there is a better response available to substitute for it. The better responses come from processes of hypothetical reasoning and cognitive simulation that are a unique aspect of Type 2 processing. When we reason hypothetically, we create temporary models of the world and test out action. Then we must be able to prevent our representations of the real world from becoming confused with representations of imaginary situations.
Figure2.1 A preliminary dual-process model
Toward a tripartite model
A simple task of letter pronunciation might entail encoding the letter, storing it in short-term memory, comparing it with information stored in long-term memory, and, if a match occurs, making a response decision and then executing a motor response. We will refer to this level of analysis as the “algorithmic level”.
But if we turn to an analysis of goals, desires and beliefs to understand a situation, we will refer to the “reflective mind”.
Figure2.2 The tripartite structure of the mind
Fig 2.2 identifies variation in fluid intelligence (Gf) with individual differences in the efficiency of processing of the algorithmic mind.
Cognitive ability and thinking dispositions
The differences between the 2 minds are capture in another well-established distinction in the measurement of individual differences: the distinction between cognitive ability and the thinking dispositions. The thinking dispositions concern a person’s goals and goal hierarchy (need for cognition, consideration of future consequences, need for closure, dogmatism…).
In fact the cognitive abilities assessed by intelligence tests are not of this type. They are not about high level personal goals and their regulation, or the tendency to change beliefs in the face of contrary evidence or about how knowledge acquisition is internally regulated when not externally directed. The actual measures of intelligence in use (IQ) assess only algorithmic-level-cognitive capacity.
Fig 2.2 identifies variation in fluid intelligence (Gf) with individual differences in the efficiency of processing of the algorithmic mind. To put it simply, the concept of rationality encompasses 2 things: thinking dispositions and algorithmic-level capacity. Rational thinking depends on our thinking dispositions as well as on our algorithmic efficiency. Moreover IQ tests certainly involve System 2, but they force the use of Type 2 processing.
Thus, as long as variation in thinking dispositions is not perfectly correlated with variation in fluid intelligence, there is a statistical possibility that rationality and intelligence diverge. In fact, substantial empirical evidence indicates that individual differences in thinking dispositions and intelligence are far from perfectly correlated.
For example, one does not maximize the deliberativeness dimension because such a person might get lost in interminable pondering and never make a decision.
Figure2.3 Model of the tripartite structure of the mind (more complete)
A kind of type 2 processing (serial associative cognition) does not involve fully explicit cognitive simulation.
Limit of Type 1 processes
Recall that the category of type 1 processes is composed of:
- affective responses
- previously learned responses that have been practiced to automaticity
- conditioned responses
- adaptive modules that have been shaped by our evolutionary history
These cover many situations indeed, but modern life still creates many problems for which none of these mechanisms is suited.
The Wason’s four card selection task (1966: letter/number K – A – 8 – 5)) is a good example where System 1 and 2 are abused.
“Consider 4 cards lying on a table. Each one of the cards has a letter on one side and a number on the other side. Here is a rule: if a card has a vowel on its letter side, then it has an even number on its number side. 2 of the cards are letter-side up, and 2 are number-side up. Your task is to decide which card or cards must be turned over in order to find out whether the rule is true or false. The 4 cards confronting the subject have the stimuli K, A, 8 and 5 showing.
The correct answer is A and 5, the only 2 cards that could show the rule to be false. However the majority of subjects incorrectly answer A and 8, showing a so-called matching bias.
In this example subjects relied often on serial associative cognition rather than exhaustive simulation of an alternative world. A world that includes situations in which the rule is false.
People are cognitive misers because their basic tendency is to default to processing mechanisms of low computational expense.
Dual-process theory has heretofore highlighted only Rule 1 of the cognitive miser. Analysis by Type 1 processing is done whenever possible. But defaulting to Type 1 processing is not always possible, particularly in novel situations. Type 2 processing procedures will be necessary in such cases, but a cognitive miser default is operating even there. When Type 2 processing is necessary, default to serial associative cognition with a focal bias (not fully decoupled cognitive simulation).
Then a third function of the reflective mind is necessary (arrow F of Fig 2.3), initiating an interrupt of serial associative cognition.
Overcoming miserly processing (detection, override and mindware*)
[* The mental knowledge and procedures that a person uses to solve problems or make decisions]
Heuristics and biases tasks were designed for human brains, not animal brains. It has been designed for brains that could at least potentially experience conscious conflict (detection). The abilities of Type 2 processing enable a cognitive critique of our beliefs and our desires.
The computational power available to sustain the serial simulation is indexed by fluid intelligence, whereas the power of the cultural tools (knowledge) used during serial simulation is in part indexed by differences in crystallized intelligence.
Don’t misinterpret. We do not mean that rationality cannot be assessed in nonhuman animals.
Heuristics and biases
In heuristics and biases tasks, the subject must detect the inadequacy of the type 1 response, in order to request and use Type 2 processing to both suppress the Type 1 response and to simulate a better alternative.
In fact, the 3 mental intertwined characteristics required for rational thinking are:
- The necessity of overriding Type 1 processing must be detected. The subject must be able to detect a conflict between his or her intuitive response to a problem and the response dictated by learned normative rules.
- The mindware that allows the computation of more rational responses needs to be available and accessible during simulation activities
- Third, algorithmic level cognitive capacity is needed so that override and simulation activities can be sustained.
Taxonomy of thinking errors
When approaching any problem, our brains have available various computational mechanisms for dealing with the situation. Some mechanisms have great computational power (they can solve a large number of novel problems). However, these mechanisms take up a great deal of attention, tend to be slow and tend to interfere with other thoughts.
Human are cognitive misers because their basic tendency is to default to processing mechanisms of low computational expense.
5 types of reasoning errors
- Necessity for override is not detected (no Type 2 processing is done at all)
- Subject attempts to override Type 1 processing but fails. Psychologists and economists often term these situations as reflecting problems of self-control. But it comprises more than just willpower and self-control. For example, the Linda conjunction problem (Kahneman 1983) is the quintessential dual-process conflict.
- The 3rd type of miserly processing represents a tendency to overeconomize during type 2 processing
- The 4th type represents a response with the intuition generated by the system 1.
- Last category is the contaminated mindware. In short, a subject may hold many specific clusters of misinformation that would make his or her behavior less rational.
Some examples of miserly processing
A bat and a ball cost $1.10 in total. The bat costs $1 more than the ball. How much does the ball cost?
When answering this problem, many people give the first response that comes to mind – 10 cents. In this case the bat would then have to cost $1.10 with a total cost of $1.20. These subjects certainly do not explicitly consider 5 cents as an alternative.
Another common paradigm used to assess belief bias in syllogistic reasoning.
- All flowers have petals
- All roses have petals
- Therefore, all roses are flowers.
The sentence used as conclusion is valid, but the syllogism is thus invalid. The processing struggle described here represents a classic case of override failure. “All flowers” does not exclude other things with petals.
The framing effects represent the classic example of serial association cognition. People respond differently depending on different wordings of an equivalent problem. Take, for example, the fact that many people are willing to pay more for hamburger meat described as 94 percent fat free than they would have been willing to pay for the same meat described as containing 6 percent fat.
Framework for classifying the types of rational thinking tasks
- Probabilistic and statistical reasoning (Bayesian Reasoning, Base rate neglect, Conjunction Fallacy, Gambler’s Fallacy, Failure to use sample size information, Probability matching).
- Scientific reasoning (How science converges on an explanation, Hypothesis testing and Falsifiability, covariation detection paradigms)
- Avoidance of miserly information processing (Disjunctive Reasoning, Ratio Bias, Syllogism, Intuition)
- Irrelevant context effect in decision making (Framing effect, Anchoring effect, preference anomalies)
- Myside bias (argument evaluation)
- Rational Temporal discounting
- Overconfidence (Knowledge calibration)
- Probabilistic numeracy, financial literacy, sensitivity to expected value and risk knowledge
- Contaminated mindware (Superstitious Thinking, Anti Science Attitude, Conspiracy Beliefs, dysfunctional Personal Beliefs)
- Thinking dispositions (Open-minded thinking, Deliberative, Future Orientation, differentiation of emotions)
Measure scales (RQ) for dispositions and attitudes of rationality
Perceptual speed, discrimination accuracy, working memory capacity, and the efficiency of the retrieval of information stored in long-term memory are examples of cognitive capacities.
Thinking dispositions, in contrast, are better viewed as cognitive style. Reflectivity and belief flexibility are “good” cognitive styles only in the sense that most people are too low on both dimension.
Open mind has similarities to measures such as need for closure measures (Kruglanski 1996), dogmatism scale (Rokeach, 1960), belief in simple knowledge (Schommer 1990).
Deliberative thinking corresponds to the scales of need for cognition and typical intellectual engagement measures (Goff 1992).
(*) In French
Drawing on theoretical work and empirical research from the last 2 decades, the authors present the first prototype for an assessment of rational thinking (RQ) analogous to the IQ test. They called it the CART (Comprehensive Assessment of Rational thinking).
The CART contains many additional measures of rational thinking in comparison with 2 other broad measures of general rational thinking (A-DMC – Parker 2005, HCTA – Halpern 2008). It has been meticulously tested with over 4,000 participants. The result is the first genuine ‘gold standard’ of rational thinking even if certain CART subtests depend on the Euro-American cultural context.
Subjects sometimes took as long as 3 hours to complete the full-form (20 subtests vs the 11 subtests of the short-form).
Access to the CART Test or equivalent is under copyright material. Unfortunately, there is no free access
Modernity increasingly requires skills of judgment and decision making
By constructing a formal assessment device, the RQ measures the skills that IQ testing has largely ignored. The RQ explicitly taps knowledge related to rational action and belief, and taps the reflective mind that an IQ test does not or few.
It is true that heuristics and biases problems seem more hostile than typical IQ test problem. IQ tests assess the algorithmic power of the mind in benign environments. But IQ tests do not pick up these hostile aspects of the cognitive environment of modernity.
The system 1 often gives us a useful first approximation to the optimal response in a given situation, and it does so without stressing cognitive capacity. But modernity increasingly requires decontextualization which is not the strength of the System 1.
The skills of judgment and decision making are cognitive skills that are the foundation of rational thought and action, and they are missing from intelligence tests (IQ …).