In 2009, David Bourget and David Chalmers ran the PhilPapers Survey (results, paper), sending questions to “all regular faculty members” at top “Ph.D.-granting [philosophy] departments in English-speaking countries” plus ten other philosophy departments deemed to have “strength in analytic philosophy comparable to the other 89 departments”.
Bourget and Chalmers now have a new PhilPapers Survey out, run in 2020 (results, paper). I’ll use this post to pick out some findings I found interesting. Keep in mind that I’m an opinionated LessWrong contributor focusing on topics and results that dovetail with things I’m curious about (e.g., ‘why do academic decision theorists and LessWrong decision theorists disagree so much?’), not giving a neutral overview of the whole 100-question survey.
The new survey’s target population consists of:
(1) in Australia, Canada, Ireland, New Zealand, the UK, and the US: all regular faculty members (tenuretrack or permanent) in BA-granting philosophy departments with four or more members (according to the PhilPeople database); and (2) in all other countries: English-publishing philosophers in BA-granting philosophy departments with four or more English-publishing faculty members.
In order to make comparisons to the 2009 results, the 2020 survey also looked at a “2009-comparable departments” list selected using similar criteria to the 2009 survey:
It should be noted that the “2009-comparable department” group differs systematically from the broader target population in a number of respects. Demographically, it includes a higher proportion of UK-based philosophers and analytic-tradition philosophers than the target population. Philosophically, it includes a lower proportion of theists, along with many other differences evident in comparing 2020 results in table 1 (all departments) to table 9 (2009-comparable departments).
Based on this description, I expect the “2009-comparable departments” in the 2020 survey to be more elite, influential, and reasonable than the 2020 “target group”, so I mostly focus on 2009-comparable departments below. In the tables below, if the row doesn’t say “Target” (i.e., target group), the population is “2009-comparable departments”.
Note that in the 2020 survey (unlike 2009), respondents could endorse multiple answers.
1. Decision theory
Newcomb’s problem: The following groups (with n noting their size, and skipping people who skipped the question or said they weren’t sufficiently familiar with it) endorsed the following options in the 2020 survey:
Group
n
one box
two boxes
diff
Philosophers (Target)
1071
31%
39%
8%
Decision theorists (Target)
48
21%
73%
58%
Philosophers
470
28%
43%
15%
Decision theorists
22
23%
73%
50%
5% of decision theorists said they “accept a combination of views”, and 9% said they were “agnostic/undecided”.
I think decision theorists are astonishingly wrong here, so I was curious to see if other philosophy fields did better.
I looked at every field where enough surveyed people gave their views on Newcomb’s problem. Here they are in order of ‘how much more likely are they to two-box than to one-box’:
Group
n
one box
two boxes
diff
Philosophers of gender, race, and sexuality
13
23%
54%
31%
20th-century-philosophy specialists
23
13%
43%
30%
Social and political philosophers
62
19%
48%
29%
Philosophers of law
14
14%
43%
29%
Phil. of computing and information (Target)
18
22%
50%
28%
Philosophers of social science
24
33%
58%
25%
Philosophers of biology
20
30%
55%
25%
General philosophers of science
53
26%
51%
25%
Philosophers of language
100
25%
47%
22%
Philosophers of mind
96
25%
45%
20%
Philosophers of computing and information
5
20%
40%
20%
19th-century-philosophy specialists
10
10%
30%
20%
Philosophers of action
32
28%
47%
19%
Logic and philosophy of logic
58
26%
43%
17%
Philosophers of physical science
25
28%
44%
16%
Epistemologists
129
32%
47%
15%
Meta-ethicists
72
29%
43%
14%
Metaphysicians
123
32%
44%
12%
Normative ethicists
102
32%
42%
10%
Philosophers of religion
15
33%
40%
7%
Philosophers of mathematics
19
32%
37%
5%
17th/18th-century-philosophy specialists
39
31%
36%
5%
Metaphilosophers
23
35%
39%
4%
Applied ethicists
50
34%
38%
4%
Philosophers of cognitive science
50
32%
36%
4%
Aestheticians
14
29%
21%
-8%
Greek and Roman philosophy specialists
17
41%
29%
-12%
(Note that many of these groups are small-n. Since philosophers of computing and information were an especially small and weird group, and I expect LWers to be extra interested in this group, I also looked at the target-group version for this field.)
Every field did much better than decision theory (by the “getting more utility in Newcomb’s problem” metric). However, the only fields that favored one-boxing over two-boxing was ancient Greek and Roman philosophy, and aesthetics.
After those two fields, the best fields were philosophy of cognitive science, applied ethics, metaphilosophy, philosophy of mathematics, and 17th/18th century philosophy (only 4-5% more likely to two-box than one-box), followed by philosophy of religion, normative ethics, and metaphysics.
My quick post-hoc, low-confidence guess about why these fields did relatively well is (hiding behind a spoiler tag so others can make their own unanchored guesses):
My inclination is to model the aestheticians, historians of philosophy, philosophers of religion, and applied ethicists as ‘in-between’ analytic philosophers and the general public (who one-box more often than they two-box, unlike analytic philosophers). I think of specialists in those fields as relatively normal people, who have had less exposure to analytic-philosophy culture and ideas and whose views therefore tend to more closely resemble the views of some person on the street.
This would also explain why the “2009-comparable departments”, who I expected to be more elite and analytic-philosophy-ish, did so much worse than the “target group” here.
I would have guessed, however, that philosophers of gender/race/sexuality would also have done relatively well on Newcomb’s problem, if ‘analytic-philosophy-ness’ were the driving factor.
I’m pretty confused about this, though the small n for some of these populations means that a lot of this could be pretty random. (E.g., network effects: a single just-for-fun faculty email thread about Newcomb’s problem could convince a bunch of philosophers of sexuality that two-boxing is great. Then this would show up in the survey because very few philosophers of sexuality have ever even heard of Newcomb’s problem, and the ones who haven’t heard of it aren’t included.)
At the same time, my inclination is to treat philosophers of cognitive science, mathematics, normative ethics, metaphysics, and metaphilosophy as ‘heavily embedded in analytic philosophy land, but smart enough (/ healthy enough as a field) to see through the bad arguments for two-boxing to some extent’.
There’s also a question of why cognitive science would help philosophers do better on Newcomb’s problem, when computer science doesn’t. I wonder if the kinds of debates that are popular in computer science are the sort that attract people with bad epistemics? (‘Wow, the Chinese room argument is amazing, I want to work in this field!’) I really have no idea, and wouldn’t have predicted this in advance.
Normative ethics also surprises me here. And both of my explanations for ‘why did field X do well?’ are post-hoc, and based on my prior sense that some of these fields are much smarter and more reasonable than others.
It’s very plausible that there’s some difference between the factors that make aestheticians one-box more, and the factors that make philosophers of cognitive science one-box more. To be confident in my particular explanations, however, we’d want to run various tests and look at various other comparisons between the groups.
The fields that did the worst after decision theory were philosophy of gender/race/sexuality, 20th-century philosophy, philosophy of language, philosophy of law, political philosophy, and philosophy of biology, of social science, and of science-in-general.
A separate question is whether academic decision theory has gotten better since the 2009 survey. Eyeballing the (small-n) numbers, the answer is that it seems to have gotten worse: two-boxing became even more popular (in 2009-comparable departments), and one-boxing even less popular:
n=31 for the 2009 side of the comparison, n=22 for the 2020 side. The numbers above are different from the ones I originally presented because Bourget and Chalmers include “skip” and “insufficiently familiar” answers, and exclude responses that chose multiple options, in order to make the methodology more closely match that of the 2009 survey.
2. (Non-animal) ethics
Regarding “Meta-ethics: moral realism or moral anti-realism?”:
Group
n
moral realism
moral anti-realism
Philosophers (Target)
1719
62%
26%
Philosophers
630
62%
27%
Applied ethicists
64
67%
23%
Normative ethicists
132
74%
16%
Meta-ethicists
94
68%
22%
Regarding “Moral judgment: non-cognitivism or cognitivism?”:
Group
n
cognitivism
non-cognitivism
Philosophers (Target)
1636
69%
21%
Philosophers
594
70%
20%
Applied ethicists
62
76%
23%
Normative ethicists
132
82%
14%
Meta-ethicists
93
76%
16%
Regarding “Morality: expressivism, naturalist realism, constructivism, error theory, or non-naturalism?”:
Group
n
non-nat
nat realism
construct
express
error
Philosophers (Target)
1024
27%
32%
21%
11%
5%
Philosophers
386
25%
33%
19%
10%
5%
Applied ethicists
40
20%
35%
38%
5%
0%
Normative ethicists
90
34%
36%
21%
8%
1%
Meta-ethicists
68
37%
29%
18%
13%
7%
Regarding “Normative ethics: virtue ethics, consequentialism, or deontology?” (putting in parentheses the percentage that only chose the option in question):
Group
n
deontology
consequentialism
virtue ethics
combination
All (Target)
1741
32% (20%)
31% (21%)
37% (25%)
16%
All
631
37% (23%)
32% (22%)
31% (19%)
17%
Applied...
63
56% (33%)
38% (22%)
33% (8%)
27%
Normative...
132
46% (27%)
32% (20%)
36% (17%)
25%
Meta...
94
44% (28%)
32% (22%)
24% (13%)
19%
Excluding responses that endorsed multiple options, we can see that normative ethicists have moved away from deontology and towards virtue ethics since 2009, though deontology is still the most popular:
30 normative-ethicist respondents also wrote in “pluralism” or “pluralist” in the 2020 survey.
Regarding “Trolley problem (five straight ahead, one on side track, turn requires switching, what ought one do?): don’t switch or switch?”:
Group
n
switch
don’t switch
Philosophers (Target)
1736
63%
13%
Philosophers
635
68%
12%
Applied ethicists
63
71%
13%
Normative ethicists
132
70%
13%
Meta-ethicists
94
70%
13%
Regarding “Footbridge (pushing man off bridge will save five on track below, what ought one do?): push or don’t push?”:
Regarding “Well-being: hedonism/experientialism, desire satisfaction, or objective list?”:
Group
n
hedonism
desire satisfaction
objective list
Philosophers (Target)
967
10%
19%
53%
Philosophers
348
9%
19%
54%
Applied ethicists
43
16%
26%
56%
Normative ethicists
90
11%
21%
63%
Meta-ethicists
62
11%
24%
55%
Moral internalism “holds that a person cannot sincerely make a moral judgment without being motivated at least to some degree to abide by her judgment”. Regarding “Moral motivation: externalism or internalism?”:
Group
n
internalism
externalism
Philosophers (Target)
1429
41%
39%
Philosophers
528
38%
42%
Applied ethicists
57
53%
37%
Normative ethicists
128
34%
51%
Meta-ethicists
92
35%
47%
One of the largest changes in philosophers’ views since the 2009 survey is that philosophers have somewhat shifted toward externalism. In 2009, internalism was 5% more popular than externalism; now externalism is 3% more popular than internalism.
(Again, the 2009-2020 comparisons give different numbers for 2020 in order to make the two surveys’ methodologies more similar.)
Philosophers of computing and information (Target)
18
44%
50%
Philosophers
366
58%
33%
Philosophers of cognitive science
40
48%
50%
Philosophers of mind
91
62%
34%
Regarding “Mind: non-physicalism or physicalism?”:
Group
n
physicalism
non-physicalism
Philosophers (Target)
1733
52%
32%
Phil of computing and information (Target)
25
64%
24%
Philosophers
630
59%
27%
Philosophers of cognitive science
73
78%
12%
Philosophers of computing and information
5
60%
20%
Philosophers of mind
135
60%
26%
Regarding “Consciousness: identity theory, panpsychism, eliminativism, dualism, or functionalism?”:
Group
n
dualism
elim
function
identity
panpsychism
Philosophers (Target)
1020
22%
5%
33%
13%
8%
Phil of computing (Target)
16
25%
19%
31%
0%
0%
Philosophers
362
22%
4%
33%
14%
7%
Phil of cognitive science
43
14%
12%
40%
19%
2%
Philosophers of mind
95
24%
2%
38%
12%
10%
Regarding “Zombies: conceivable but not metaphysically possible, metaphysically possible, or inconceivable?” (also noting “agnostic/undecided” results):
Group
n
inconceivable
conceivable +impossible
possible
agnostic
Philosophers (Target)
1610
16%
37%
24%
11%
Phil of computing (Target)
24
29%
25%
17%
8%
Philosophers
582
15%
42%
22%
11%
Phil of cognitive science
72
24%
50%
11%
6%
Phil of computing
5
20%
40%
20%
0%
Philosophers of mind
132
20%
51%
17%
3%
My understanding is that the “psychological view” of personal identity more or less says ‘you’re software’, the “biological view” says ‘you’re hardware’, and the “further-fact view” says ‘you’re a supernatural soul’. Regarding “Personal identity: further-fact view, psychological view, or biological view?”:
Group
n
biological
psychological
further-fact
Philosophers (Target)
1615
19%
44%
15%
Phil of computing… (Target)
23
22%
70%
4%
Philosophers
598
20%
44%
13%
Philosophers of cognitive science
69
26%
55%
6%
Philosophers of computing...
5
40%
60%
20%
Philosophers of mind
130
26%
47%
12%
Comparing this to some other philosophy subfields, as a gauge of their health:
Group
n
biological
psychological
further-fact
Decision theorists
18
17%
67%
17%
Epistemologists
144
22%
35%
17%
General philosophers of science
58
29%
55%
7%
Metaphysicians
150
23%
39%
15%
Normative ethicists
127
14%
46%
15%
Philosophers of language
120
18%
38%
17%
Philosophers of mathematics
18
17%
61%
6%
Philosophers of religion
25
24%
16%
44%
Decision theorists come out of this looking pretty great (I claim). This is particularly interesting to me, because some people diagnose the ‘academic decision theorist vs. LW decision theorist’ disagreement as coming down to ‘do you identify with your algorithm or with your physical body?’.
The above is some evidence that either this diagnosis is wrong, or academic decision theorists haven’t fully followed their psychological view of personal identity to its logical conclusions.
Regarding “Mind uploading (brain replaced by digital emulation): survival or death?” (adding answers for “the question is too unclear to answer” and “there is no fact of the matter”):
Group
n
survival
death
Q too unclear
no fact
Philosophers (Target)
1016
27%
54%
5%
4%
Phil of computing… (Target)
19
47%
42%
11%
0%
Philosophers
369
28%
54%
5%
4%
Decision theorists
12
42%
8%
8%
17%
Philosophers of cognitive science
37
35%
51%
5%
3%
Philosophers of mind
91
34%
52%
4%
2%
From my perspective, decision theorists do great on this question — very few endorse “death”, and a lot endorse “there is no fact of the matter” (which, along with “survival”, strike me as good indirect signs of clear thinking given that this is a kind-of-terminological question and, depending on terminology, “death” is at best a technically-true-but-misleading answer).
Also, a respectable 25% of decision theorists say “agnostic/undecided”, which is almost always something I give philosophers points for — no one’s an expert on everything, a lot of these questions are confusing, and recognizing the limits of your own understanding is a very positive sign.
Regarding “Chinese room: doesn’t understand or understands?” (adding “the question is too unclear to answer” responses):
Group
n
understands
doesn’t
Q too unclear
Philosophers (Target)
1031
18%
67%
6%
Phil of computing… (Target)
18
22%
56%
17%
Philosophers
381
18%
66%
6%
Philosophers of cognitive science
44
34%
50%
7%
Philosophers of mind
91
15%
70%
8%
Regarding “Other minds (for which groups are some members conscious?)” (looking only at the “2009-comparable departments”, except for philosophy of computing and information because there aren’t viewable results for that subgroup):
(Options: adult humans; cats; fish; flies; worms; plants; particles; newborn babies; current AI systems; future AI systems.)
(Respondent groups: philosophers; applied ethicists; decision theorists; meta-ethicists; metaphysicians; normative ethicists; philosophy of biology; philosophers of cognitive science; philosophers of computing and information; philosophers of mathematics; philosophers of mind.)
n
adult h
cat
fish
fly
worm
plant
partic
baby h
AI
AI fut
Phil
404
97%
93%
68%
36%
24%
7%
2%
89%
2%
43%
App
35
100%
94%
77%
31%
17%
3%
0%
89%
0%
63%
Dec
14
86%
86%
71%
36%
21%
0%
0%
64%
0%
57%
MtE
64
98%
92%
72%
38%
23%
6%
2%
91%
0%
47%
MtP
108
98%
97%
71%
42%
32%
7%
3%
95%
4%
49%
Nor
85
99%
94%
73%
35%
22%
5%
0%
88%
4%
41%
Bio
13
100%
85%
62%
38%
15%
8%
0%
69%
8%
54%
Cog
44
98%
98%
73%
39%
25%
11%
2%
95%
2%
50%
Com
19
100%
89%
68%
32%
26%
5%
0%
89%
11%
58%
Mat
14
100%
93%
86%
57%
43%
7%
0%
93%
7%
50%
Min
92
99%
97%
79%
45%
30%
10%
4%
96%
0%
43%
I am confused, delighted, and a little frightened that an equal (and not-super-large) number of decision theorists think adult humans and cats are conscious. (Though as always, small n.)
Also impressed that they gave a low probability to newborn humans being conscious — it seems hard to be confident about the answer to this, and being willing to entertain ‘well, maybe not’ seems like a strong sign of epistemic humility beating out motivated reasoning.
Also, 11% of philosophers of cognitive science think PLANTS are conscious??? Friendship with philosophers of cognitive science ended, decision theorists new best friend.
Regarding “Eating animals and animal products (is it permissible to eat animals and/or animal products in ordinary circumstances?): vegetarianism (no and yes), veganism (no and no), or omnivorism (yes and yes)?”:
Group
n
omnivorism
vegetarianism
veganism
Philosophers (Target)
1764
48%
26%
18%
Philosophers
643
46%
27%
21%
Applied ethicists
64
33%
23%
41%
Normative ethicists
131
40%
27%
31%
Meta-ethicists
94
41%
23%
24%
Philosophers of mind
134
48%
31%
12%
Philosophers of cognitive science
73
48%
29%
14%
4. Metaphysics, philosophy of physics, and anthropics
Regarding “Sleeping beauty (woken once if heads, woken twice if tails, credence in heads on waking?): one-half or one-third?” (including the answers “this question is too unclear to answer,” “accept an alternative view,” “there is no fact of the matter,” and “agnostic/undecided”):
Group
n
1⁄3
1⁄2
unclear
alt
no fact
agnostic
Philosophers (Target)
429
28%
19%
8%
1%
3%
40%
Philosophers
191
27%
20%
4%
2%
5%
42%
Decision theorists
13
54%
8%
0%
15%
0%
23%
Epistemologists
70
33%
20%
3%
3%
1%
40%
Logicians and phil of logic
28
36%
14%
4%
7%
4%
36%
Phil of cognitive science
18
28%
22%
6%
0%
0%
44%
Philosophers of mathematics
6
50%
17%
0%
0%
0%
33%
Regarding “Cosmological fine-tuning (what explains it?): no fine-tuning, brute fact, design, or multiverse?”:
Group
n
design
multiverse
brute fact
no fine-tuning
Philosophers (Target)
807
17%
15%
32%
22%
Philosophers
289
14%
16%
35%
22%
Decision theorists
13
8%
23%
23%
38%
General phil of science
33
9%
21%
48%
24%
Metaphysicians
93
28%
18%
32%
13%
Phil of cognitive science...
32
6%
19%
50%
13%
Phil of physical science
16
13%
25%
38%
6%
Regarding “Quantum mechanics: epistemic, hidden-variables, many-worlds, or collapse?”:
What is the metaphysical basis for causal connection? That is, what is the difference between causally related and causally unrelated sequences?
The question of connection occupies the bulk of the vast literature on causation. [...] Fortunately, the details of these many and various accounts may be postponed here, as they tend to be variations on two basic themes. In practice, the nomological, statistical, counterfactual, and agential accounts tend to converge in the indeterministic case. All understand connection in terms of probability: causing is making more likely. The change, energy, process, and transference accounts converge in treating connection in terms of process: causing is physical producing. Thus a large part of the controversy over connection may, in practice, be reduced to the question of whether connection is a matter of probability or process (Section 2.1).
Regarding “Causation: process/production, primitive, counterfactual/difference-making, or nonexistent?”:
Group
n
counterfactual
process
primitive
non
Philosophers (Target)
892
37%
23%
21%
4%
Philosophers
342
39%
22%
21%
3%
Decision theorists
14
71%
7%
7%
7%
Metaphysicians
103
32%
28%
28%
5%
Philosophers of cognitive science
38
58%
21%
13%
5%
Philosophers of physical science
16
63%
19%
0%
6%
Regarding Foundations of mathematics: constructivism/intuitionism, structuralism, set-theoretic, logicism, or formalism?:
Group
n
constructiv
formal
logicism
structural
set
Philosophers (Target)
600
15%
6%
12%
21%
15%
Philosophers
229
12%
4%
12%
24%
17%
Philosophers of mathematics
15
13%
7%
7%
40%
33%
5. Superstition
Regarding “God: atheism or theism?” (with subfields ordered by percentage that answered “theism”):
Group
n
theism
Philosophy (Target)
1770
19%
Philosophy
645
13%
Philosophy of religion
27
74%
Medieval and Renaissance philosophy
10
60%
Philosophy of action
42
21%
17th/18th century philosophy
64
20%
20th century philosophy
36
19%
Metaphysics
153
16%
Normative ethics
132
16%
19th century philosophy
22
14%
Asian philosophy
7
14%
Decision theory
22
14%
Philosophy of mind
135
13%
Aesthetics
30
13%
Ancient Greek and Roman philosophy
31
13%
Applied ethics
64
13%
Epistemology
153
13%
Logic and philosophy of logic
70
13%
Philosophy of language
128
12%
Philosophy of social science
27
11%
Meta-ethics
94
10%
Philosophy of law
21
10%
Philosophy of mathematics
20
10%
Social and political philosophy
100
10%
Philosophy of gender, race, and sexuality
23
9%
Philosophy of biology
24
8%
Philosophy of physical science
26
8%
General philosophy of science
65
6%
Philosophy of cognitive science
73
5%
Philosophy of computing and information
5
0%
Philosophy of the Americas
5
0%
Continental philosophy
15
0%
Philosophy of computing and information (Target)
25
0%
Metaphilosophy
29
0%
This question is ‘philosophy in easy mode’, so seems like a decent proxy for field health / competence (though the anti-religiosity of Marxism is a confounding factor in my eyes, for fields where Marx is influential).
The “A-theory of time” says that there is a unique objectively real “present”, corresponding to “which time seems to me to be right now”, that is universal and observer-independent, contrary to special relativity. The “B-theory of time” says that there is no such objective, universal “present”.
This provides another good “reasonableness / basic science literacy” litmus test, so I’ll order the subfields (where enough people in the field answered at all) by how much more they endorse B-theory over A-theory. Regarding “Time: B-theory or A-theory?”:
Group
n
A-theory
B-theory
diff
Philosophy (Target)
1123
27%
38%
11%
Philosophy
449
22%
44%
22%
19th century philosophy
13
31%
8%
-23%
Philosophy of religion
22
45%
27%
-18%
Medieval and Renaissance philosophy
9
33%
22%
-11%
Philosophy of law
10
30%
20%
-10%
Aesthetics
18
22%
17%
-5%
Philosophy of social science
17
24%
24%
0%
Social and political philosophy
42
29%
33%
4%
Philosophy of gender, race, and sexuality
12
25%
33%
8%
Ancient Greek and Roman philosophy
21
24%
33%
9%
Philosophy of action
31
29%
39%
10%
20th century philosophy
22
23%
36%
13%
Normative ethics
78
31%
44%
13%
Philosophy of cognitive science
49
18%
31%
13%
Meta-ethics
60
27%
43%
16%
Philosophy of mathematics
17
12%
29%
17%
Asian philosophy
5
20%
40%
20%
Applied ethics
29
31%
52%
21%
Epistemology
113
21%
42%
21%
Philosophy of mind
111
20%
41%
21%
17th/18th century philosophy
44
14%
36%
22%
Metaphilosophy
23
17%
39%
22%
Metaphysics
144
28%
51%
23%
Logic and philosophy of logic
61
28%
52%
24%
Phil of computing and information (Target)
20
25%
50%
25%
Philosophy of language
107
22%
54%
32%
General philosophy of science
48
19%
56%
37%
Philosophy of biology
16
19%
63%
44%
Decision theory
16
13%
63%
50%
Philosophy of physical science
26
12%
62%
50%
Decision theorists doing especially well here is surprising to me! Especially since they didn’t excel on theism; if they’d hit both out of the park, from my perspective that would have been a straightforward update to “wow, decision theorists are really exceptionally reasonable as analytic philosophers go, even if they’re getting Newcomb’s problem in particular wrong”.
As is, this still strikes me as a reason to be more optimistic that we might be able to converge with working decision theorists in the future. (Or perhaps more so, a reason to be relatively optimistic about persuading decision theorists vs. people working in most other philosophy areas.)
(Added: OK, after writing this I saw decision theorists do great on the ‘personal identity’ and ‘mind uploading’ questions, and am feeling much more confident that productive dialogue is possible. I’ve added those two questions earlier in this post.)
(Added added: OK, decision theorists are also unusually great on “which things are conscious?” and they apparently love MWI. How have we not converged more???)
Another interesting result is “Philosophical methods (which methods are the most useful/important?)”, which finds (looking at analogous-to-2009 departments):
66% of philosophers think “conceptual analysis” is especially important, 14% disagree.
60% say “empirical philosophy”, 12% disagree.
59% say “formal philosophy”, 10% disagree.
51% say “intuition-based philosophy”, 27% disagree.
44% say “linguistic philosophy”, 23% disagree.
39% say “conceptual engineering”, 23% disagree.
29% say “experimental philosophy”, 39% disagree.
8. How have philosophers’ views changed since 2009?
Bourget and Chalmers’ paper has a table for the largest changes in philosophers’ views since 2009:
As noted earlier in this post, one of the larger shifts in philosophers’ views was a move away from moral internalism and toward externalism.
On ‘which do you endorse, classical logic or non-classical?’ (a strange question, but maybe this is something like ‘what kind of logic is reality’s source code written in?’), non-classical logic is roughly as unpopular as ever, but fewer now endorse classical logic, and more give answers like “insufficiently familiar with the issue” and “the question is too unclear to answer”:
Epistemic contextualism says that the accuracy of your claim that someone “knows” something depends partly on contextual features — e.g., the standards for “knowledge” can rise “as the stakes rise or the skeptical doubts become more serious”.
Here, it was the less popular view (invariantism) that lost favor; and the view that lost favor again lost it via an increase in ‘other’ answers (especially “insufficiently familiar with the issue” and “agnostic/undecided”) more so than increased favor for its rival view (contextualism):
Humeanism (a misnomer, since Hume himself wasn’t a Humean, though his skeptical arguments helped inspire the Humeans) say that “laws of nature” aren’t fundamentally different from other observed regularities, they’re just patterns that humans have given a fancy high-falutin name to; whereas anti-Humeans think there’s something deeper about laws of nature, that they in some sense ‘necessitate’ things to go one way rather than another.
(Maybe Humeans = ‘laws of nature are program outputs like any other’, non-Humeans = ‘laws of nature are part of reality’s source code’?)
Once again, one view lost favor (the more popular view, non-Humeanism), but the other didn’t gain favor; instead, more people endorsed “insufficiently familiar with the issue”, and “agnostic/undecided”, etc.:
Philosophers in 2020 are more likely to say that “yes”, humans have a priori knowledge of some things (already very much the dominant view):
‘Aesthetic value is objective’ was favored over ‘subjective’ (by 3%) in 2009; now ‘subjective’ is favored over ‘objective’ (by 4%). “Agnostic/undecided” also gained ground.
Philosophers mostly endorsed “switch” in the trolley dilemma, and still do; but “don’t switch” gained a bit of ground, and “insufficiently familiar with the issue” lost ground.
Moral realism also became a bit more popular (was endorsed by 56% of philosophers, now 60%), as did compatibilism about free will (was 59% compatibilism, 14% libertarianism, 12% no free will; now 62%, 13%. and 10%).
The paper also looked at the individual respondents who answered the survey in both 2009 and 2020. Individuals tended to update away from switching in the trolley dilemma, away from consequentialism, and toward virtue ethics and non-cognitivism. They also updated toward Platonism about abstract objects, and away from ‘no free will’.
These are all comparisons across 2009-target-population philosophers in general, however. In most (though not all) cases, I’m more interested in the views of subfields specialized in investigating and debating a topic, and how the subfield’s view changes over time. Hence my earlier sections largely focused on particular fields of philosophy.
2020 PhilPapers Survey Results
In 2009, David Bourget and David Chalmers ran the PhilPapers Survey (results, paper), sending questions to “all regular faculty members” at top “Ph.D.-granting [philosophy] departments in English-speaking countries” plus ten other philosophy departments deemed to have “strength in analytic philosophy comparable to the other 89 departments”.
Bourget and Chalmers now have a new PhilPapers Survey out, run in 2020 (results, paper). I’ll use this post to pick out some findings I found interesting. Keep in mind that I’m an opinionated LessWrong contributor focusing on topics and results that dovetail with things I’m curious about (e.g., ‘why do academic decision theorists and LessWrong decision theorists disagree so much?’), not giving a neutral overview of the whole 100-question survey.
The new survey’s target population consists of:
In order to make comparisons to the 2009 results, the 2020 survey also looked at a “2009-comparable departments” list selected using similar criteria to the 2009 survey:
Based on this description, I expect the “2009-comparable departments” in the 2020 survey to be more elite, influential, and reasonable than the 2020 “target group”, so I mostly focus on 2009-comparable departments below. In the tables below, if the row doesn’t say “Target” (i.e., target group), the population is “2009-comparable departments”.
Note that in the 2020 survey (unlike 2009), respondents could endorse multiple answers.
1. Decision theory
Newcomb’s problem: The following groups (with n noting their size, and skipping people who skipped the question or said they weren’t sufficiently familiar with it) endorsed the following options in the 2020 survey:
5% of decision theorists said they “accept a combination of views”, and 9% said they were “agnostic/undecided”.
I think decision theorists are astonishingly wrong here, so I was curious to see if other philosophy fields did better.
I looked at every field where enough surveyed people gave their views on Newcomb’s problem. Here they are in order of ‘how much more likely are they to two-box than to one-box’:
(Note that many of these groups are small-n. Since philosophers of computing and information were an especially small and weird group, and I expect LWers to be extra interested in this group, I also looked at the target-group version for this field.)
Every field did much better than decision theory (by the “getting more utility in Newcomb’s problem” metric). However, the only fields that favored one-boxing over two-boxing was ancient Greek and Roman philosophy, and aesthetics.
After those two fields, the best fields were philosophy of cognitive science, applied ethics, metaphilosophy, philosophy of mathematics, and 17th/18th century philosophy (only 4-5% more likely to two-box than one-box), followed by philosophy of religion, normative ethics, and metaphysics.
My quick post-hoc, low-confidence guess about why these fields did relatively well is (hiding behind a spoiler tag so others can make their own unanchored guesses):
My inclination is to model the aestheticians, historians of philosophy, philosophers of religion, and applied ethicists as ‘in-between’ analytic philosophers and the general public (who one-box more often than they two-box, unlike analytic philosophers). I think of specialists in those fields as relatively normal people, who have had less exposure to analytic-philosophy culture and ideas and whose views therefore tend to more closely resemble the views of some person on the street.
This would also explain why the “2009-comparable departments”, who I expected to be more elite and analytic-philosophy-ish, did so much worse than the “target group” here.
I would have guessed, however, that philosophers of gender/race/sexuality would also have done relatively well on Newcomb’s problem, if ‘analytic-philosophy-ness’ were the driving factor.
I’m pretty confused about this, though the small n for some of these populations means that a lot of this could be pretty random. (E.g., network effects: a single just-for-fun faculty email thread about Newcomb’s problem could convince a bunch of philosophers of sexuality that two-boxing is great. Then this would show up in the survey because very few philosophers of sexuality have ever even heard of Newcomb’s problem, and the ones who haven’t heard of it aren’t included.)
At the same time, my inclination is to treat philosophers of cognitive science, mathematics, normative ethics, metaphysics, and metaphilosophy as ‘heavily embedded in analytic philosophy land, but smart enough (/ healthy enough as a field) to see through the bad arguments for two-boxing to some extent’.
There’s also a question of why cognitive science would help philosophers do better on Newcomb’s problem, when computer science doesn’t. I wonder if the kinds of debates that are popular in computer science are the sort that attract people with bad epistemics? (‘Wow, the Chinese room argument is amazing, I want to work in this field!’) I really have no idea, and wouldn’t have predicted this in advance.
Normative ethics also surprises me here. And both of my explanations for ‘why did field X do well?’ are post-hoc, and based on my prior sense that some of these fields are much smarter and more reasonable than others.
It’s very plausible that there’s some difference between the factors that make aestheticians one-box more, and the factors that make philosophers of cognitive science one-box more. To be confident in my particular explanations, however, we’d want to run various tests and look at various other comparisons between the groups.
The fields that did the worst after decision theory were philosophy of gender/race/sexuality, 20th-century philosophy, philosophy of language, philosophy of law, political philosophy, and philosophy of biology, of social science, and of science-in-general.
A separate question is whether academic decision theory has gotten better since the 2009 survey. Eyeballing the (small-n) numbers, the answer is that it seems to have gotten worse: two-boxing became even more popular (in 2009-comparable departments), and one-boxing even less popular:
n=31 for the 2009 side of the comparison, n=22 for the 2020 side. The numbers above are different from the ones I originally presented because Bourget and Chalmers include “skip” and “insufficiently familiar” answers, and exclude responses that chose multiple options, in order to make the methodology more closely match that of the 2009 survey.
2. (Non-animal) ethics
Regarding “Meta-ethics: moral realism or moral anti-realism?”:
Regarding “Moral judgment: non-cognitivism or cognitivism?”:
Regarding “Morality: expressivism, naturalist realism, constructivism, error theory, or non-naturalism?”:
Regarding “Normative ethics: virtue ethics, consequentialism, or deontology?” (putting in parentheses the percentage that only chose the option in question):
Excluding responses that endorsed multiple options, we can see that normative ethicists have moved away from deontology and towards virtue ethics since 2009, though deontology is still the most popular:
30 normative-ethicist respondents also wrote in “pluralism” or “pluralist” in the 2020 survey.
Regarding “Trolley problem (five straight ahead, one on side track, turn requires switching, what ought one do?): don’t switch or switch?”:
Regarding “Footbridge (pushing man off bridge will save five on track below, what ought one do?): push or don’t push?”:
Regarding “Human genetic engineering: permissible or impermissible?”:
Regarding “Well-being: hedonism/experientialism, desire satisfaction, or objective list?”:
Moral internalism “holds that a person cannot sincerely make a moral judgment without being motivated at least to some degree to abide by her judgment”. Regarding “Moral motivation: externalism or internalism?”:
One of the largest changes in philosophers’ views since the 2009 survey is that philosophers have somewhat shifted toward externalism. In 2009, internalism was 5% more popular than externalism; now externalism is 3% more popular than internalism.
(Again, the 2009-2020 comparisons give different numbers for 2020 in order to make the two surveys’ methodologies more similar.)
3. Minds and animal ethics
Regarding “Hard problem of consciousness (is there one?): no or yes?”:
Regarding “Mind: non-physicalism or physicalism?”:
Regarding “Consciousness: identity theory, panpsychism, eliminativism, dualism, or functionalism?”:
Regarding “Zombies: conceivable but not metaphysically possible, metaphysically possible, or inconceivable?” (also noting “agnostic/undecided” results):
+impossible
My understanding is that the “psychological view” of personal identity more or less says ‘you’re software’, the “biological view” says ‘you’re hardware’, and the “further-fact view” says ‘you’re a supernatural soul’. Regarding “Personal identity: further-fact view, psychological view, or biological view?”:
Comparing this to some other philosophy subfields, as a gauge of their health:
Decision theorists come out of this looking pretty great (I claim). This is particularly interesting to me, because some people diagnose the ‘academic decision theorist vs. LW decision theorist’ disagreement as coming down to ‘do you identify with your algorithm or with your physical body?’.
The above is some evidence that either this diagnosis is wrong, or academic decision theorists haven’t fully followed their psychological view of personal identity to its logical conclusions.
Regarding “Mind uploading (brain replaced by digital emulation): survival or death?” (adding answers for “the question is too unclear to answer” and “there is no fact of the matter”):
From my perspective, decision theorists do great on this question — very few endorse “death”, and a lot endorse “there is no fact of the matter” (which, along with “survival”, strike me as good indirect signs of clear thinking given that this is a kind-of-terminological question and, depending on terminology, “death” is at best a technically-true-but-misleading answer).
Also, a respectable 25% of decision theorists say “agnostic/undecided”, which is almost always something I give philosophers points for — no one’s an expert on everything, a lot of these questions are confusing, and recognizing the limits of your own understanding is a very positive sign.
Regarding “Chinese room: doesn’t understand or understands?” (adding “the question is too unclear to answer” responses):
Regarding “Other minds (for which groups are some members conscious?)” (looking only at the “2009-comparable departments”, except for philosophy of computing and information because there aren’t viewable results for that subgroup):
(Options: adult humans; cats; fish; flies; worms; plants; particles; newborn babies; current AI systems; future AI systems.)
(Respondent groups: philosophers; applied ethicists; decision theorists; meta-ethicists; metaphysicians; normative ethicists; philosophy of biology; philosophers of cognitive science; philosophers of computing and information; philosophers of mathematics; philosophers of mind.)
I am confused, delighted, and a little frightened that an equal (and not-super-large) number of decision theorists think adult humans and cats are conscious. (Though as always, small n.)
Also impressed that they gave a low probability to newborn humans being conscious — it seems hard to be confident about the answer to this, and being willing to entertain ‘well, maybe not’ seems like a strong sign of epistemic humility beating out motivated reasoning.
Also, 11% of philosophers of cognitive science think PLANTS are conscious??? Friendship with philosophers of cognitive science ended, decision theorists new best friend.
Regarding “Eating animals and animal products (is it permissible to eat animals and/or animal products in ordinary circumstances?): vegetarianism (no and yes), veganism (no and no), or omnivorism (yes and yes)?”:
4. Metaphysics, philosophy of physics, and anthropics
Regarding “Sleeping beauty (woken once if heads, woken twice if tails, credence in heads on waking?): one-half or one-third?” (including the answers “this question is too unclear to answer,” “accept an alternative view,” “there is no fact of the matter,” and “agnostic/undecided”):
Regarding “Cosmological fine-tuning (what explains it?): no fine-tuning, brute fact, design, or multiverse?”:
Regarding “Quantum mechanics: epistemic, hidden-variables, many-worlds, or collapse?”:
From SEP:
Regarding “Causation: process/production, primitive, counterfactual/difference-making, or nonexistent?”:
Regarding Foundations of mathematics: constructivism/intuitionism, structuralism, set-theoretic, logicism, or formalism?:
5. Superstition
Regarding “God: atheism or theism?” (with subfields ordered by percentage that answered “theism”):
This question is ‘philosophy in easy mode’, so seems like a decent proxy for field health / competence (though the anti-religiosity of Marxism is a confounding factor in my eyes, for fields where Marx is influential).
The “A-theory of time” says that there is a unique objectively real “present”, corresponding to “which time seems to me to be right now”, that is universal and observer-independent, contrary to special relativity. The “B-theory of time” says that there is no such objective, universal “present”.
This provides another good “reasonableness / basic science literacy” litmus test, so I’ll order the subfields (where enough people in the field answered at all) by how much more they endorse B-theory over A-theory. Regarding “Time: B-theory or A-theory?”:
Decision theorists doing especially well here is surprising to me! Especially since they didn’t excel on theism; if they’d hit both out of the park, from my perspective that would have been a straightforward update to “wow, decision theorists are really exceptionally reasonable as analytic philosophers go, even if they’re getting Newcomb’s problem in particular wrong”.
As is, this still strikes me as a reason to be more optimistic that we might be able to converge with working decision theorists in the future. (Or perhaps more so, a reason to be relatively optimistic about persuading decision theorists vs. people working in most other philosophy areas.)
(Added: OK, after writing this I saw decision theorists do great on the ‘personal identity’ and ‘mind uploading’ questions, and am feeling much more confident that productive dialogue is possible. I’ve added those two questions earlier in this post.)
(Added added: OK, decision theorists are also unusually great on “which things are conscious?” and they apparently love MWI. How have we not converged more???)
6. Identity politics topics
Regarding “Race: social, unreal, or biological?”:
(Note that many respondents said ‘yes’ to multiple options.)
7. Metaphilosophy
Regarding “Philosophical progress (is there any?): a little, a lot, or none?”:
Regarding “Philosophical knowledge (is there any?): a little, none, or a lot?”:
Another interesting result is “Philosophical methods (which methods are the most useful/important?)”, which finds (looking at analogous-to-2009 departments):
66% of philosophers think “conceptual analysis” is especially important, 14% disagree.
60% say “empirical philosophy”, 12% disagree.
59% say “formal philosophy”, 10% disagree.
51% say “intuition-based philosophy”, 27% disagree.
44% say “linguistic philosophy”, 23% disagree.
39% say “conceptual engineering”, 23% disagree.
29% say “experimental philosophy”, 39% disagree.
8. How have philosophers’ views changed since 2009?
Bourget and Chalmers’ paper has a table for the largest changes in philosophers’ views since 2009:
As noted earlier in this post, one of the larger shifts in philosophers’ views was a move away from moral internalism and toward externalism.
On ‘which do you endorse, classical logic or non-classical?’ (a strange question, but maybe this is something like ‘what kind of logic is reality’s source code written in?’), non-classical logic is roughly as unpopular as ever, but fewer now endorse classical logic, and more give answers like “insufficiently familiar with the issue” and “the question is too unclear to answer”:
Epistemic contextualism says that the accuracy of your claim that someone “knows” something depends partly on contextual features — e.g., the standards for “knowledge” can rise “as the stakes rise or the skeptical doubts become more serious”.
Here, it was the less popular view (invariantism) that lost favor; and the view that lost favor again lost it via an increase in ‘other’ answers (especially “insufficiently familiar with the issue” and “agnostic/undecided”) more so than increased favor for its rival view (contextualism):
Humeanism (a misnomer, since Hume himself wasn’t a Humean, though his skeptical arguments helped inspire the Humeans) say that “laws of nature” aren’t fundamentally different from other observed regularities, they’re just patterns that humans have given a fancy high-falutin name to; whereas anti-Humeans think there’s something deeper about laws of nature, that they in some sense ‘necessitate’ things to go one way rather than another.
(Maybe Humeans = ‘laws of nature are program outputs like any other’, non-Humeans = ‘laws of nature are part of reality’s source code’?)
Once again, one view lost favor (the more popular view, non-Humeanism), but the other didn’t gain favor; instead, more people endorsed “insufficiently familiar with the issue”, and “agnostic/undecided”, etc.:
Philosophers in 2020 are more likely to say that “yes”, humans have a priori knowledge of some things (already very much the dominant view):
‘Aesthetic value is objective’ was favored over ‘subjective’ (by 3%) in 2009; now ‘subjective’ is favored over ‘objective’ (by 4%). “Agnostic/undecided” also gained ground.
Philosophers mostly endorsed “switch” in the trolley dilemma, and still do; but “don’t switch” gained a bit of ground, and “insufficiently familiar with the issue” lost ground.
Moral realism also became a bit more popular (was endorsed by 56% of philosophers, now 60%), as did compatibilism about free will (was 59% compatibilism, 14% libertarianism, 12% no free will; now 62%, 13%. and 10%).
The paper also looked at the individual respondents who answered the survey in both 2009 and 2020. Individuals tended to update away from switching in the trolley dilemma, away from consequentialism, and toward virtue ethics and non-cognitivism. They also updated toward Platonism about abstract objects, and away from ‘no free will’.
These are all comparisons across 2009-target-population philosophers in general, however. In most (though not all) cases, I’m more interested in the views of subfields specialized in investigating and debating a topic, and how the subfield’s view changes over time. Hence my earlier sections largely focused on particular fields of philosophy.