,
104 tweets,
34 min read
Read on Twitter
Ok about to kick off the Rethinking Formal Methods in Philosophy #RFMP2019 conference, gonna livetweet it in this thread today. First talk by Joshua Knobe, discussing what Formal Methods are Most Used in Philosophy? - empirical evidence about what we're actually doing!
Why care what formal methods we should actually use? Knobe says he is thinking in particular about what we should teach students, maybe especially graduate students given that they will go on to shape the field #RFMP2019
He used to think this was just a bureaucratic question about how we organise our departments - but he now thinks it is important that we think what sort of knowledge will be best for our students as they go on in life, and how we answer this changes what knowledge they will have.
Why wonder about what sort of formal methods philosophers actually use? Partly because folk should know what methods we actually use to gonna engage with. But Knobe cautions that we may reasonably decide that folk dont know what they ought, so shouldn't just follow status quo.
Compare and contrast psychology grad programmes - they require some knowledge of statistics, because that is the form of mathematics most often used in psychology, even though there are other formal theories (e.g. game theory) that are sometimes used in psych papers.
Does something parallel hold in philosophy, where we can require a particular kind of formal theory because it most often used, even though one sometimes sees other methods used? #RFMP2019
Just as a Knobe looked at what's published in Philosophical Studies in 2016-2017 (375 papers), hand coding all papers as to what particular kinda formal method they used if any. He's keen to stress that he's aware that this is a very rough method! #RFMP2019
What does he mean by using formal methods? Can't count if you are just mentioning formal results, or using symbols for brevity ("consider action A that Phi's...") but Knobe
also didn't require that you are making a novel contribution to mathematics. #RFMP2012
also didn't require that you are making a novel contribution to mathematics. #RFMP2012
E.g. paper on epistemic modals that constructs a logical model, or shows that two claims are equivalent by means of symbolic transformations - this would count as using logic, even if the model is not especially original or the equivalence shown is not unknown to mathematicians
Comparison of what number of papers in his sample used logic versus other formal theories - note some papers appear in both groups #RFMP2019 
A break down of what the non logic papers were doing - Knobe notes that all of these are methods that make heavy use of probability theory. #RFMP2019
This is relevant to philosophy pedagogy, since plausibly a course which shed light on one of them would help a student understand many of the others. #RFMP2019
The vast majority of logic papers, didn't make use of non-logic methods as well (those that did tended to all use things from formal epistemology or statistics). #RFMP2019
Of the kind of logic folk use in their papers in his sample, it tends much more to be the kind of logic that makes use of the type of logic done in foundations of mathematics as opposed to tools more often used by linguists or folk trying to understand natural language #RFMP2019
(Knobe is very worried that here his sample is misleading - Phil Studies isn't where people are going with more linguisticsy flavoured logical work, and he thinks that younger philosophers are disproportionately those involved in formal semantics and that isn't accounted for.)
Taking for now the results to be representative, Knobe thinks that we ought expand what formal requirements we require of graduate students, to allow them to keep up with the revealed diversity of uses we see. #RFMP2019
You might worry that we shouldn't be so presentist, even if rnow we're using this great variety if one wants to understand past philosophy this would give us a different focus. Knobe agrees, but thinks we should also neglect the future - what will help students shape the future?
Another worry - we don't want to just stick to the actual practice of philosophy, we can have good reason to depart from what has or is been/being done. He agrees, but wants to stress that we shouldn't *not* allow students to use common formal methods to meet their requirement.
Analogy to ancient philosophy: philosophers learn lots of Aristotle but rarely learn about Zhuangzi. In response to this he thinks we ought allow students to count courses on Zhuangzi or compelling them to take courses on Zhuangzi as well as Aristotle to meet ancient requirement.
Joshua Knobe said he would be open to working more on this topic about what formal methods are used in philosophy to do something more rigorous, so if you got the time and the skills and the desire, maybe get in contact!
Ok now we're moving on Helen De Cruz' talk on Reflections on Teaching Experimental Philosophy to Undergraduates - let's go #RFMP2019 , might be of interest to @xphilosopher and co
@xphilosopher De Cruz begins by framing what we want from a humanities education with remarks from Martha Nussbuam and José Medina, It can help us engage in critical examination of oneself and one's traditions, avoiding a kind of naive status quo bias where whatever one is doing must be best
It helps one understand oneself as a member of a broader society in one's nation and beyond, De Cruz notes explicitly that this rationale would directly entail teaching more non-Western philosophy than we presently do, we're part of an interconnected globe
And it improves narrative imagination, helps us see the world in ways we don't now - she mentions a course on Nietzsche being taught at her institution which provides new and very challenging ways of understanding the world.
So how does experimental philosophy fit in to this? She thinks that it can give our undergrads both the kinda skills that many STEM graduates have, while also helping our students foster and develop those humanities skills.
De Cruz describes a course she taught in the UK, which culminated in the students replicating the Knobe effect (jstor.org/stable/pdf/438…) and also replicating a paper by Nichols claiming norms survive more if they tap into our disgust emotions (jstor.org/stable/10.1086…) #RFMP2019
She notes that her students (both informally and also in student evals) tended not to see Experimental Philosophy as all that distinct from their other courses, it was just one more tool in the general tool box that one is learning as an undergraduate philosopher. #RFMP2019
Her students did replicate the Knobe effect, as expected.
Along the way to teaching people how to do this they learned basics of what replication is and science and why it is valuable, some statistical literacy which they then applied to discussing the relationship between pre-referendum polling and the actual Brexit result. #RFMP2019
(Also some cool examples for thinking about the relationship between correlation and causation - turns out number of storks on roofs and number of children born do actually correlate! Can you think why that would be? #RFMP2019)
((Also @MorganKThomp's work just got a shout out! Experimental philosophers represent!))
@MorganKThomp One challenge De Cruz faced in teaching X-Phi was: she wanted students to learn the underlying mathematical basis for the stats, but many humanities students were not necessarily prepared for that. But on the flip side there was high attendance and students reported loving it.
@MorganKThomp Does philosophy teach transferable skills? De Cruz notes her libertarian friends seem keen to stress the high earnings and good LSAT scores etc of philosophers is a case of correlation not causation, she's not sure why correlation between libertarians and making this argument.
That said, she is not sure herself that the correlations between philosophical skills and later high scores on quantitate evaluations of employability (e.g. LSAT or MCAT scores) was all that indicative of much we can advertise in our favour.
De Cruz hopes her students learned the humility that comes with experimental work, realising the many difficulties that come with seriously trying to test your ideas against empirical data. #RFMP2019
Summary slide
In Q&A De Cruz says that there is something empowering about teaching students that they don't just have to passively take in empirically gathered statistical data, they understand how such knowledge is made and feel they could contribute to or challenge it on its own terms.
Now Eric Steinhart is talking to us about Can Mathematics Save Philosophy? He spends his life discussing with administrators the value of philosophy, trying to persuade them of our value and preserve programmes and positions. So he's going to talk about our relationship to admin.
Alas, he says, much of his tome talking to administrators about philosophy is spent looking at graphs shaped as such
He thinks more mathematicisatiion would help with things administrators care about and make our students more employable, but time spent teaching these difficult tools can detract from other things we care about, and the simplified methods we'd teach would risk becoming dogma.
But Eric none the less says that philosophers should be more polemical in their advocacy of formal methods, that mathematicising our ideas often and genuinely adds clarity, and what's more we're not doing anything legible to administrators as valuable in our current practice.
((Lol this talk is spicy please don't shoot the messenger twitter I am just live tweeting! #RFMP2019))
"Philosophers are often not critical about philosophy. We're critical about other philosophers, but not the field itself... Maybe we shouldn't tolerate so much" e.g. say "If you can't give me a mathematical model of what free will amounts to, you're not talking about anything!"
Steinhart also says that professional philosophy goes along with an auteur theory of creative genius, a Romantic model, leads to philosophers not collaborating. This has negative consequences, pantheons, canons and a romanticisation that gets in the way of diversifying the field.
That said, he also says he sees a lot of pseudo-mathematics, some of which he'd say goes so far as to be a deep attempt at bamboozling or hiding under false authority of mathematics, and on other occasions mathematicisation leads to pointless epicycles. #RFMP2019
Steinhart proposes that we create more on the model of formal philosophy syllabi banks that can be widely shared, along with class room tested teaching materials. This would be useful to practitioners and apparently administrators really respect that too. #RFMP2019
He thinks that a more mathematical approach to philosophy would lead to questions modelled less on the natural sciences and this would be a good idea - rather than asking "Do universals exist?" we ask "What are properties of systems which include a commitment to universals?"
More ambitiously, Steinhart proposes that more formal courses and degree structures would help us make alliances with other fields. Joint degree structures with computer sciences and engineering degrees could be intellectually fruitful and attract funding and students.
What college does philosophy sit in, where do or should we sit in the university? Does formal philosophy change the position of philosophy in the university? - questions Steinhart is asking himself and prompting us to reflect on.
Should be thinking about where do our majors go with their careers, where do we want them to go? How does formal philosophy interact with jobs for philosophy majors? These are very concrete practical questions, he says, declining student enrolment requires we take them seriously.
Serious worries he has: making philosophy more mathsy will put some students off, & will contribute to less diversity among our hires; what data he has seen suggests formal philosophy skews towards men and whites. And hiring patterns and trends may be against formal philosophy!
On latter - he says many of the practical arguments for formal philosophy also speak in favour of applied ethics focus, and in fact hiring patterns and administrative interest are starting to track that. He hopes that formal methods (e.g. statistical analysis) can help here too.
Steinhart: "I don't just wanna make administrators happy, I want to make administrators happy by doing better philosophy" #RFMP2019
Ok now we're back from lunch with Ray Briggs' Baby Logic, on how introductory logic courses can serve the goals of students who are not themselves focussed on formal work - we'll be focussed on a course Ray has actually taught, learning from their experiences. Let's go! #RFMP2019
Briggs discusses what they've seen taught in informal logic classes - fallacies (which Briggs not so keen on, introduces "debate club feel", learning gotchas), identifying argumentative structure (yay!) , and syllogisms (yep, in the sense of traditional logic! Bit old fashioned!)
Briggs & Thomas Icard teach a course at Stanford which combines elements of baby logic, other formal tools (probability, decision theory, game theory, statistics), and informal logic topics mentioned. Shout out to @bweatherson's available teaching material brian.weatherson.org/LectureNotes.h…
This broad array allows them to maintain connection to other topics of interest in philosophy, so students can see how what they’re learning here is useful to them elsewhere #RFMP2019
Briggs now going through the Allais paradox with everyone, I'm wondering whether the conference participants gonna turn out to be consistent with the classical model of decision theoretic rationality... #RFMP2019
... Briggs told everyone the answer! So I guess we'll never know if this room is filled with rational people or not.
When teaching their course their big challenges are: each section has to be quite rushed, and therefore with great urgency working out how to focus on what is most important for their broader education. And also students tend to have strong maths anxiety re statistics. #RFMP2019
Interestingly Briggs also feels the need for more publicly available textbooks, mirroring Steinhart's talk. Theme emerging? #RFMP2019
Briggs also gave a shout out to Jenn Wang, they said that Jenn did great work in developing a lot of the resources they now use in their course. Thanks Jenn! jennwang.org #RFMP2019
Briggs has mapped out the course structure, in terms of what elements build on other parts #RFMP2019
They also discuss some pedagogical things they use in the field - for more advanced students they prepare lots of optional puzzles, this is apparently very popular. They also make a point of publicly owning their mistakes at the front of the class, modelling learning from error.
Question from audience gives pedagogical defence of teaching syllogistic; nice intro to quantifiers, simple cases with no nesting, combines with Briggs' earlier point that the Venn diagramatic interpretation helps. Syllogistic is also paraconsistent so it's an intro to relevance!
One thing that makes Briggs & Icard's course fit in so well is that the Stanford undergrad body are typically quite formally inclined and so it fits into their broader study at the school. Worth tailoring your courses to the needs of your particular school environment. #RFMP2019
(Lmao a questioner mentions that my blog post on
@DailyNousEditor has already got a bunch of people angry on Facebook. Stay mad, haters! dailynous.com/2019/09/19/for…)
@DailyNousEditor has already got a bunch of people angry on Facebook. Stay mad, haters! dailynous.com/2019/09/19/for…)
@DailyNousEditor Questioner asks about how formal methods can be introduced via philosophical puzzles which aren't specifically motivated by reflection on formal systems themselves. Briggs responds with issues in philosophy of science: problem of induction & IBE may draw from probabilistic tools.
@DailyNousEditor Now we hear from Gregory Wheeler, with his talk "Too Boole for School" on what they are doing at the Frankfurt School (yes we made that joke) in building a programme with lots of connections to CS and machine learning #RFMP2019
@DailyNousEditor Wheeler opens with some discussion of this paper - he notes that for a couple of the 5 topics they highlight, but most especially in artificial intelligence, the field has largely moved on from logic based methods.
cs.cmu.edu/~rwh/papers/un…
cs.cmu.edu/~rwh/papers/un…
@DailyNousEditor He's now talking about the role of Nvidia's role in stimulating machine learning research, since the chips they came up with for image processing were used by folks in Germany and elsewhere to think about the basis of our now quite advanced methods of image recognition.
@DailyNousEditor On Wheeler's telling this has been a real problem for logical based AI, since it has nothing like this kind of success story which could generate the same level of interest as have the less logic-y machine learning methods.
@DailyNousEditor "Logical AI is dead" - The Frankfurt School #RFMP2019
@DailyNousEditor Wheeler isn't so impressed by "theory is dead" rhetoric that sometimes accompanies these advances in more black box or opaque prediction methods from machine learning, but none the less thinks its an interesting project to see how far we can take these methods. #RFMP2019
@DailyNousEditor What is more, these methods opens up or render urgent questions, some about causal discovery or decision theoretic reasoning on the basis of results from these methods, and second integrated ethics programmes that consider the consequences of the social uptake of these methods.
@DailyNousEditor Some deep conceptual questions at the intersection of logic, machine learning, and ethics - what does it mean to verify ethical software? #RFMP2019
@DailyNousEditor Wheeler "we're philosophers, we love mathematically pathological objects, that's why I got into this business -- I just love negative results!"
@DailyNousEditor In one connection to Steinhart's talk, Wheeler talks about how administrative pressure to show employability benefits to their students drives their decisions to focus on the connections between philosophy and machine learning. #RFMP2019
@DailyNousEditor There have actually been two start ups to come out of the faculty in Greg Wheeler's department - not exactly usual fare for a philosophy department! #RFMP2019
@DailyNousEditor But he also mentions that the cultural respect for philosophy in Germany has made it easier to get things going there, administration do give you some benefit of the doubt and are willing to take risks on philosophy programmes with unusual focuses, since after all its philosophy!
@DailyNousEditor None the less, he thinks the kind of programme they're trying to build is exportable beyond Germany: this is in fact a worthy and even profitable line of inquiry for philosophy and the broader world. Just a matter of getting admin to see whats in their enlightened self interest.
@DailyNousEditor Questioner: even if logical programme in AI is no longer cutting edge, philosophy in grad school might still have a place since people still need basic logic? Wheeler agrees, but says often not much more than basic is needed: also good to know paradoxes & history of the field.
Wheeler expresses some hope that by hearing about what we're doing, we can mutually inspire and encourage people to try new things out at their own schools. #RFMP2019
In response to a question Wheeler clarifies that he doesn't expect the close and fruitful relationship between logic and computer science more generally (or even all AI) to go away, even if this very successful machine learning programme is going in a different direction.
(Wheeler mentions the role of logical methods in verification for the blockchain - here at #RFMP2019 it's the first blockchain mention I've heard at a philosophy conference, so posting this video in honour of the momentous occasion )
Some discussion now of interacting with computer scientists: Wheeler's experience is machine learning folk generally are interested in conceptual work done by ethicists, but sometimes aren't aware of what's done in philosophy. Philosophers need to better advertise their work.
Someone raised the question of whether P=NP, but unfortunately we've run out of time so we can't settle it this session. #RFMP2019
Now for our last talk of the day, Conor Mayo-Wilson with a talk called "Formal Methods in Education: is TMTOWTDI?" (There More Than One Way To Do It?) #RFMP2019
Mayo-Wilson thinks all grad programmes should have a requirement that people learn enough logic and set theory to follow a proof (and not be afraid of symbolic reasoning), and enough probability to follow statistics and decision theory, then beyond that let 1000 flowers bloom.
Mayo-Wilson says that since his talk is very ambitious in how much he hopes to get done here, there'll be no argument. Let's go.
Shout out to @RrrichardZach who found all 35 ranked departments in PGR in 2004 required logic for grad students. Conor Mayo-Wilison found today 46 of "top 50" (his quotation marks) require logic today, of the other four 3 have "logic or decision theory" disjunctive requirements.
@RrrichardZach Mayo-Wilson doesn't find claims of instrumental benefits of logic persuasive: often under-evidenced, opportunity costs not considered, and students don't regularly practice skill after the classes. Instead he thinks logic is a core part of philosophy, self-justifying. #RFMP2019
@RrrichardZach He finds it more plausible that logic is required to understand key moments in 20th century analytic philosophy, but again it would be better to have some research backing this up. Also, as in Knobe's talk, we can't uncritically take status quo and reinforce it with our teaching.
@RrrichardZach "For philosophy to survive and be more meaningful, I think we ought to craft a narrative about the value of philosophy and how it can improve lives" - Conor Mayo-Wilson #RFMP2019
@RrrichardZach Mayo-Wilson wants to draw from old narratives from early modern European philosophy, weaving together new epistemology, formal and technological developments, and a new political and ethical settlement through their efforts as a philosophers. This can be a model for us. #RFMP2019
@RrrichardZach For our new(ish) epistemology let's focus on new loci of knowledge: complex social institutions including machines (here he draws on feminist, cog sci, and sociological work). We should reject armchair physical/social theory, and develop new languages for describing knowledge.
@RrrichardZach These kind of projects should guide our decisions about what goes into our *required* formal methods class. #RFMP2019
@RrrichardZach But Mayo-Wilson notes that he still thinks we ought do this in-house rather than have our students go off and take classes in maths, stats, CS, etc. Philosophers ask different questions and bring to bear more systematic thinking about important related normative issues. #RFMP2019
@RrrichardZach Gave some examples of non-philosophy textbooks making false claims about interpreting probability, and simplistic claims about measure of accuracy and what utiility is. Worrying because students often think that normative questions are already settled once formal claims are made.
@RrrichardZach Philosophers need enough formal methods to be able to critique and improve important mathematical tools used to guide and govern our public life. Similar in some sense to claims by Steinhart and De Cruz. Philosophers need formal methods to keep our critical edge. #RFMP2019
@RrrichardZach Some discussion now about whether or not we should have any required courses at all, disagreement among the audience but Mayo-Wilson at least defending the position that minimal requirements (not just formal requirements) needed to create a discipline with shared vocabulary.
@RrrichardZach Ok that's day one of the #RFMP2019 ended, hope yinz enjoyed - see you back here tomorrow!
