“Do philosophers really think differently than non-philosophers?”
That’s the question taken up by a team of researchers in a recent study published in Philosophical Psychology. Focusing on differences between philosophers and non-philosophers in regard to neurodivergent traits and responses to thought experiments, the researchers found some support, in the aggregate, for a “yes” answer to that question.
The following guest post is by Thomas Nadelhoffer (College of Charleston), and is based on the article, “Thinking Differently: Neurodivergent Traits and Responses to Thought Experiments in Philosophers and the General Population,” co-authored with Paul McKee (Duke), Khaoula Boulhadi (Southern Connecticut State), and Helena Weymouth (College of Charleston).
How Different Are Philosophers?
by Thomas Nadelhoffer
“The history of philosophy is to a great extent that of a certain clash of human temperaments [. . .] Yet, his temperament really gives him a stronger bias than any of his more strictly objective premises. It loads the evidence for him one way or the other” — William James
Neurodiversity and Philosophical Intuitions
In addition to being abstract thinkers, philosophers are stereotyped as being argumentative, overly self-absorbed in their own thoughts, and enjoy creating increasingly fine distinctions regarding everyday terms while imagining novel hypothetical situations and debating the conclusions they draw. It’s possible that many philosophers have quite distinctive personalities.
More than two millennia before the modern notion of stereotypes existed, Plato used humor to poke fun at philosophers in the Theaetetus by telling a story of Thales contemplating the stars, so engrossed that he forgot where he was walking and fell down a well. However, do philosophers really think differently than non-philosophers?
This was one of the questions behind the paper we recently published in Philosophical Psychology. We surveyed 1,254 individuals, 632 with formal training in philosophy and 622 members of the general public, and explored whether philosophers exhibit significant cognitive differences from non-philosophers using multiple measures of neurodiverse traits. To get enough responses from philosophers, we recruited participants here at Daily Nous and on Leiter Reports. We recruited the general population participants from Prolific. We preregistered our experimental design, research questions, and hypotheses.
Philosophical Temperament?
Recently, experimental philosophers and psychologists began investigating philosophical temperament, that is, the study of relationships between psychological traits and philosophically relevant judgment. Several studies have linked various personality traits to judgments concerning issues like free will, moral responsibility, intentional action, and metaethics. Research has also demonstrated that although trained to reason about certain things rationally, philosophers are susceptible to psychological factors influencing their philosophical intuitions. Our aim was to investigate this area further.
Rather than continuing to focus on personality characteristics that researchers typically measure using personality surveys (e.g., The “Big Five”), we decided to examine characteristics that relate to the following neurodiverse traits:[1] autism; ADHD; aphantasia (low visual imagery); anendophasia (i.e. low levels of “inner speech”); and representational manipulation (the ability to mentally manipulate and modify representations).[2]
Of the four significant differences found amongst participants from the general population and those with some degree of formal training in philosophy:
- Philosophers exhibited higher levels of ADHD-related traits (moderate effect size, d = .5).
- Philosophers exhibited lower levels of visual imagery (substantial effect size, d = -.6).
- Philosophers exhibited higher levels of internal speech (smaller effect size, d = .34).
- Philosophers demonstrated higher levels of representational manipulation (substantial effect size, d = .76).
All four differences found were very stable, even after we limited the criterion for “philosopher” as only participants with PhD degrees or postdoctoral appointments in philosophy.
Therefore, our findings suggest that, as a group, philosophers have a distinct cognitive profile (relative to the general population). This profile consists of higher levels of internal verbalization, enhanced capacities for mentally manipulating internal representations, lower levels of visual imagery and higher levels of ADHD traits.
Philosophers and Thought Experiments
All participants completed two classical versions of this type of experiment: the “bystander at the switch” version of the trolley problem and the “rollback universe” deterministic universe.
We also assessed how often each group misinterpreted determinism, through either bypassing (treating an agent’s beliefs, desires, and deliberation as having no causal relevance to their actions) or indeterminate intrusion (allowing multiple possible outcomes while maintaining identical initial conditions and laws).
As before, there were significant differences between philosophers and non-philosophers.
In the trolley case, philosophers were significantly more likely to believe it is acceptable to sacrifice the one person but not obligatory and less likely to believe that it is unacceptable.
In short, there was a small yet statistically reliable correlation between being a philosopher and intuitions about the bystander at the switch scenario.
However, there were larger differences in the responses to the determinism case. Approximately 50% of philosophers believed that agents in the deterministic universe have free will. In comparison, approximately 37% of members of the general public reported the same intuition. For moral responsibility, approximately 69% of philosophers and 54% of the general population judged that agents in the rollback universe are responsible. Moreover, philosophers made the bypassing error much less frequently than did members of the general public. While 17% of philosophers made this error, 63% of those in the general public made it. Similarly, while approximately 25% of philosophers made the intrusion error, approximately 33% of members of the general public made it.
Our next goal was to assess if there existed a connection between the cognitive characteristics of our subjects and their responses to philosophical thought experiments. We found that psychological traits are correlated with responses to thought experiments independently of the subject’s exposure to philosophy. Visual imagery provides a good illustration. Those who reported lower levels of visualization were more likely to make utilitarian judgments in response to the bystander at the switch scenario. Conversely, those who reported more vivid visual imagery were less likely to indicate that flipping the switch was permissible but not obligatory.
Furthermore, visual imagery influenced judgments regarding determinism. Those who indicated higher levels of visual imagery were less likely to assign free will and moral responsibility to agents in the rollback universe. Additionally, individuals with lower visualization scores were more susceptible to both bypassing and intrusion errors.[3]
Machine Learning and Sorting Philosophers and Non-Philosophers
The final analysis that we ran involved using various machine learning algorithms to see whether we could predict whether we could accurately distinguish between philosopher and non-philosophers using neurodiverse traits, responses to thought experiments, and both taken together. So, we trained various machine learning models to determine if they could categorize participants into either philosophers or non-philosophers.
For the first round of analysis, we provided the models with only participants’ answers to philosophical thought experiments. This resulted in reasonable success. The top performing model achieved approximately 78% accuracy. The strongest predictor variable was whether participants committed the bypassing error when reasoning about determinism.
Next, we provided the models with only participants’ cognitive-trait profile. We found that neurodiverse traits were more predictive than responses to thought experiments when it came to whether a participant was a philosopher or non-philosopher. The strongest predictor traits were representational manipulation, visual imagery, and ADHD symptoms. Using just these three traits, the classification accuracy was 81%. In short, if we choose someone at random from our study and we want to determine whether that person has studied philosophy, we could just use responses to the scales for measuring visualization, representational manipulation, and AHD to make reliable predictions that are quite significantly above chance.
Lastly, we asked the models to use both types of information (that is, responses to thought experiments and neurodiverse traits. By combing the two, the classification-accuracy of the models went up to 86%. The three best predictors for this mixed model were scores on bypassing, representational-manipulation, and visual imagery.
In closing, I just want to point out that our study had several limitations (e.g., it relied on self-report) and it also left several unanswered questions (e.g., we can’t speak to whether philosophy attracts people with certain cognitive styles or whether philosophical training influences cognitive style). You can find our discussion of these limitations (and more detailed discussions of our findings) in our full article. But because Justin was kind enough to post our call for participants on Daily Nous and many readers took the time to complete the survey, I thought I would share the results.
[1] It’s worth emphasizing that we were examining traits (measured via scales), not attempting to diagnose participants with autism spectrum disorders, ADHD, aphantasia (“mind blindness”), anendophasia (no internal dialogue), etc.
[2] This trait is not normally considered to be a neurodiverse trait; however, scores on the scale measuring this construct are significantly correlated with several of the neurodiverse traits listed above.
[3] We also found that other cognitive characteristics similarly correlated with philosophical judgments. However, due to space constraints, I will refrain from detailing them here (those interested can check out the paper).
ADHD often manifests differently in men & women. What percentage of your population of philosophers were men, were the controls similarly matched, and was ADHD self-reported, or based on the symptoms more likely to manifest in one sex rather than the other?
Footnote 1 in the post says “It’s worth emphasizing that we were examining traits (measured via scales), not attempting to diagnose participants with autism spectrum disorders, ADHD, aphantasia (“mind blindness”), anendophasia (no internal dialogue), etc.”
Unless I’ve misunderstood, this means that ADHD is not rendered a “trait” in this study. So these concerns about differences are about the interpretation of the results beyond the scope of this research.
Deborah, Thanks for the questions. “Someone Else” already answered part of your question. I wanted to fill in more of the details.
First, we used an abbreviated version of the following scale for measuring ADHD: ADHD Self-Report Scale (ASRS; Kessler et al., 2005, p. 18 items). Sample item: “How often do you have difficulty waiting your turn in situations when turn-taking is required?” Scale anchors: 1 = Never − 5 = Very often. α = .92.”
Second, here is what we said about gender in the limitations section of the paper: “Sixth, there is a large gender disparity between our two conditions. While 65% of participants in the general population sample were female, only 22% of those in the philosophy sample identified as female. Given this asymmetry, concerned readers may be apprehensive that the associations we found between neurodivergent traits and philosophers could partly reflect the fact that philosophers are much more likely to be men, and men are more likely
to have some of the traits we explored (e.g., autism and ADHD). However, when we conducted a follow-up analysis on a gender-balanced subsample to rule out gender as a confounding variable, the results held (see Supplemental Tables S3-45)).
Specifically, we randomly selected equal numbers of male and female participants within each group of the whole dataset (philosophers and general population), matching group sizes to the smallest available gender cell (i.e., the minimum number of males or females present in either group). Additionally, with our machine learning pipeline, gender alone predicted group membership with an accuracy of 68%, and combining gender with age yielded an accuracy of 70%. These results suggest that while demographic factors are predictive in our dataset, reflecting the gender asymmetry, they do not account for the bulk of the predictive power, underscoring the robust association between cognitive traits and philosophical
identity.”
I hope that helps.
“We found that neurodiverse traits were more predictive than responses to thought experiments when it came to whether a participant was a philosopher or non-philosopher.”
Out of context, this is just too good.
Yes, I agree with what you say here! 🙂
I’d love to see a comparison with mathematicians.
I was thinking it would definitely be helpful to have a “control group” of non-philosopher academics, to see if they have the rate of misunderstandings of non-academics (because they haven’t studied these thought experiments, or practiced this kind of careful attention to details of a thought experiment) while still having the same “neurodiverse” traits of the philosophers.
Kenny, when we preregistered this study, we originally planned to use the same paradigm to study neurodiverse traits in psychologists vs. the general population (and psychologists vs. philosophers). We had difficulty determining how to recruit enough psychologists to complete the survey. Getting philosophers to take the survey was easier because there are two widely-read philosophy blogs (Daily Nous and Leiter Reports).
To our knowledge, psychologists don’t have similar community blogs for their field. Another limitation involved the practical constraint of funding. Recruiting the 600+ participants on Prolific ate up most of our research budget (given the study’s length, it was somewhat expensive to recruit them). I paid for this study with my departmental research budget (which is very meager), and it was insufficient to conduct the follow-up study with psychologists. But you’re right that it would be interesting to use academics more generally as a control condition.
Given the popularity of psychology as a major, there were a number of participants in the general population who majored in psychology, but there weren’t enough to enable us to run a proper comparison between those participants and the participants in the philosophy condition.
Holger, the research on autism and STEM fields is partly what motivated our project. Here is what we said about these empirical findings:
“As we mentioned earlier, several studies have shown that neurodivergent traits are associated with enhanced capabilities, not just impairments. For instance, Simon Baron-Cohen and colleagues have shown that while autistic
people may have difficulties with certain forms of empathy, they tend to be high in “systematization” – i.e., they often have a “drive to analyze and/or build a system (of any kind) based on identifying input-operation-output
rules” (Baron-Cohen et al., 1997, 1998, 1999, 2001, 2003, 2007; Lawson et al., 2004). As a result, autistic individuals are particularly well-suited to work in fields where systematizing is an essential skill – e.g., architecture, biology, computer science, economics, engineering, finance, information technology, mathematics, music, and physics (Wei et al., 2013). While there is empirical evidence linking autism to both an interest in and an enhanced aptitude for STEM fields, we are not aware of any empirical evidence regarding the relationship, if any, between neurodivergent
traits and interest in, and/or enhanced aptitude for, philosophy.”
We expected similar findings in our study. Here is our hypothesis and prediction about autistic traits in philosophers and the general population:
“H1a: Given that autistic individuals have higher systematizing traits and tendencies than the general population, and philosophers often exhibit these traits and tendencies, we predicted that philosophers would score higher than the general population on the measures used to assess autism.”
The findings were mixed:
“Scores on the shortened ASQ were significantly higher among philosophy participants (M = 2.68, SD = 0.54) compared to the general population (M = 2.59, SD = 0.54), t(720.921) = 2.182, p = .029, 95% CI [0.008, 0.156], d = 0.152. However, the effect size was small, indicating a weak association between group membership and ASQ scores.
No significant difference was found on RAADS-14,(philosophers: M = 2.10, SD = 0.80; general population: M = 2.15, SD = 0.72, t(768.568) = −0.846, p = .398, 95% CI [−0.148, 0.059], d = −0.058. These results were
replicated with the full data but not the PhD+ subset. With that comparison, there was no difference with the short autism measure, but philosophers had significantly lower autism scores in the full inventory.”
Of interest: https://psychiatryonline.org/doi/10.1176/pn.39.11.0390036 Not sure how good the study is.
https://pmc.ncbi.nlm.nih.gov/articles/PMC2784305/
Also this to link it back.
How do we know that someone who commits the bypassing “error” doesn’t simply believe that, e.g., every world in which they intend other than they actually do is nevertheless one in which they act as they actually do? Given this belief, determinism would entail bypassing, at least in the sense that any world compatible with the subject’s beliefs would be one in which bypassing is true if determinism is true. (Note that I’m understanding bypassing as fatalism rather than epiphenomenalism, as the authors do, and simplifying things by abstracting away from initial conditions and laws and the like. The former seems more accurate given the description of bypassing; the latter is immaterial to my point.)
If participants made the assumption you mentioned, that, too, would be a comprehension error. In recent work with Samuel Murray, we developed measures for three errors: bypassing, indeterministic intrusion, and fatalism. For the neurodiversity study, we used measures only for the first two. But if someone views the deterministic universe based on the line of reasoning you sketched out, then they misunderstood the implications of the deterministic universe, since determinism does not entail fatalism (in the same way that determinism does not entail epiphenomenalism). The only reason we didn’t include the fatalism items is that we were already giving participants a battery of scales, so we were concerned about participant fatigue–which can be a confound. But you’re right that some participants likely mistook determinism for fatalism, which we found in my earlier work with Murray.
Isn’t fatalism just determinism + necessary (instead of contingent) laws of nature?
People who believe laws of nature are necessary may not consider the distinction between determinism and fatalism apt.
I’m not saying that determinism entails fatalism; by itself, it does not. I’m saying that it does so *when it is conjoined with certain other propositions*, and it might be that subjects believe those propositions, holding them fixed when reasoning about determinism. Then it wouldn’t seem irrational at all for them to infer fatalism, so that the label “error” is misplaced (as it is in many of these “what else do ‘the folk’ not understand” studies).
For what it’s worth: I understand fatalism, in the philosophical sense, to be the claim that certain future events would have occurred whatever the conditions preceding them. Epiphenomenalism strikes me as a species of fatalism in which the relevant conditions are phenomenally conscious states and the events are human behaviors; bypassing, as defined here and elsewhere, sounds like it could include either epiphenomenalism or fatalism restricted to conditions like intentions and desires—conscious or otherwise.
In any case, my question remains: how do we know that subjects aren’t just sneaking in confounding background beliefs which justify the entailment, rather than misunderstanding the meaning of determinism?
unless i’m mistaken, epiphenomenalists don’t deny that mental states counterfactually depend on the physical world. They just deny that they make a difference to what happens in the physical world. So most epiphenomenalists wouldn’t claim that “every world in which they intend other than they actually do is nevertheless one in which they act as they actually do”.
I’m not sure I follow. Epiphenomenalists wouldn’t claim this, I agree (except to the extent that an intention is also a phenomenal state). But then this is one of the immaterial differences I mentioned; replace “intention” with “phenomenal state” and the line you quote *is* one the epiphenomenalist would endorse—it changes from an account of fatalism about intention to fatalism about consciousness (that is, epiphenomenalism).
(If it helps: I’m assuming that an action is something which “happens in the physical world.”)
(Also, I’m assuming that the counterfactual here is equivalent to, or at least follows from, the causal inefficacy of a given mental state—phenomenal or otherwise. I think the dependence you describe *is* endorsed by the epiphenomenalist, but it’s not the one I’m describing. “Phenomenal states depend on physical states” is not the same thing as “later physical states, like actions, depend on phenomenal states.”)
No, I don’t think epiphenomanalists would endorse it with phenomenal states either. A world in which they had acted otherwise is quite likely a world in which their physical brain would have given rise to different phenomenal states (states which themselves are causally inefficacious)
I haven’t said anything about a world in which they had acted otherwise. I have described worlds in which they act the same as they actually do, but with different mental states preceding them.
Ohh but maybe you meant that a world in which they have different *phenomenal states* would be one in which they have different brain states (given the dependence of the phenomenal on the physical), and these different brain states could in turn cause different acts? In which case, epiphenomenalists *would* deny the following: given phenomenal states other than those they actually have, they would have acted the same as they actually do. And they could do that without sacrificing the causal inefficacy of phenomenal states which defines their epiphenomenalism, because the *brain states* are what explain the difference in action (rather than the phenomenal states). That might be an interesting way of showing that epiphenomenalists can’t express their (causal) view in counterfactual terms, contrary to what I suggested above.
Yes that is what I was meant, sorry if it wasn’t clear.
It is true that it means they can’t express their causal view in terms of simple counterfactual dependence but maybe the reason is the same reason why simple counterfactual dependence fails as an analysis of causation. (eg c causes both e and f, e is causally idle, but “if e hadn’t occurred, f wouldn’t have occurred). But I’m very far from being an expert here
No worries, tbh I’m losing track of the dialectic a bit (I’m not sure how this relates back to the determinism part, and I’m too lazy to think it through). But the point about epiphenomenalism (as a causal thesis) resisting counterfactual analysis is really interesting, thanks for pointing it out! Then I guess another complication is that it’s difficult to relate epiphenomenalism to both determinism and fatalism, which are naturally construed in counterfactual terms.
Haha same – all good. Glad you found it interesting! 🙂
This is really interesting. I have advanced degrees in philosophy although I never explicitly pursued it professionally (needless to say it informs everything I do nonetheless). I’m a little bit annoyed by how accurate much of this feels.
I try in my work to cultivate more empathy and I am learning to have more visual awareness and ability– but the basic wiring of thinking like a philosopher seems innate. I very clearly recall my first formal class in philosophy (although I had been reading philosophy and theology since middle school). The first discussion after a reading assignment felt like I had cognitively come home.
The low visual imagery result is really surprising. It’s surprising to me personally as I am pretty sure that I am at the opposite end of the spectrum, (which I suppose means that I am phantastic as opposed to aphantastic). I have a strong visual imagination ,enjoy sketching and painting and can call famous artworks to mind , at least enough to describe them if not to sketch them from the images in my head. The visual life of the mind (or the life of the visual mind) is very important to me. And it comes as some surprise that I am atypical in this.
But the result surprises me for another reason. I teach a course on the British moralists (roughly 1650-1800) and one the topics I emphasise is the widespread commitment on the part of British philosophers to what I call psycho-semantic empiricism, and Jonathan Bennett calls ‘concept empiricism’. This is the claim that our ideas – a broad category which includes what some would describe as ‘concepts’ – are constructed out of the copies of past sensations. We have the concept of a unicorn, for example, because we have seen horses and seen horns and can jam the images of horns onto the heads of images of horses, thus constructing the concept of a unicorn. We have no concepts that are not constructed out of past sensations and words which seem to stand for such concepts are in fact meaningless. (This last is the semantic component of psycho-semantic empiricism.) Psycho-semantic empiricism (utterly absurd as it is ) plays a big part in the philosophies of Hobbes, Locke, Berkeley and Hume with Price and Reid as prominent critics (though the thesis had been effectively pre-refuted by Descartes with his example of the chiliagon). Now it is very clear from the rhetoric of the psycho-semantic empiricists that when they talk about the impressions from which ideas are constructed, it is visual impressions that they chiefly have in mind. Indeed, I don’t think it is a doctrine that would even have occurred to people who were not strongly visual thinkers since they come close to denying that aphantastic thought is even possible. (It is notable that their opponents, Reid and Price emphasise plausible examples of aphantistic thought and aphantastic concept-formation.)
This makes me wonder whether Nadelhoffer’s result isn’t rather parochial, a generalisation that holds good for professional philosophers in the 21st century but not for philosophers considered as a temporally extended class. If so this prompts the question: Why are present-day philosophers a lot less phantastic than their eighteenth century predecessors?
A speculative answer to your final question: the logic turn at the turn of the twentieth century or the language turn in the middle of the century?
Interesting speculation. Russell was relatively aphantastic (though of course this is usually a matter of degree). He was surprised when he could visualise something (for example, his lover Ottoline’s face). I think this is connected to his aesthetic sensibility. He does not seem to have cared that much about painting or sculpture (though he was taught to have the ‘correct’ opinions by his sometime brother-in-law, Logan Pearsall Smith) though he was very sensitive to poetry. But I have no idea about the inner lives of other pioneering logicians or indeed about the inner lives of logicians in general.
There are two opposing fallacies due to generalising from ones own experience or the lack thereof. The phantastic fallacy, to which the early modern empiricists were prone, consists in denying the possibility of aphantastic thought. The aphantastic fallacy consists in denying the actuality and perhaps the possibility of phantastic thought. Two philosophers who come precious close to this are Gilbert Ryle and Dan Dennett. They both seem to deny the existence mental images in the sense that I seem to have them. For me remembering my nursery (or at least my childhood bedroom) is NOT like resembling a spectator of my nursery (whatever exactly that is supposed to mean) but exactly like spectating a resemblance of my nursery. Ryle was blissfully unaware in The Concept of Mind of how much he was only speaking for his aphantastic self.
Anyone might be startled to find Ottoline Morrell’s face suddenly intrude upon their thoughts:
https://www.npg.org.uk/collections/search/portrait/mw08219/Lady-Ottoline-Morrell
cruel!