Comments
55
I have developed a university course on existential risks and I have been teaching it at two universities: at my own university, the University of Groningen, Netherlands, for 3rd Year Liberal Arts and Sciences students, and at LMU Munich, Germany, for Philosophy as well as Logic and Philosophy of Science Master students.
Perhaps university teachers elsewhere consider introducing such a course in their curriculum. The new book "The Precipice: Existential Risk and the Future of Humanity" by Toby Ord can be used very well as a textbook for such a course. This will make it straightforward to identify suitable topics to cover and a natural order in which to tackle these topics. But which teaching formats and assessment moments should one use? Here I outline some ideas for those, hoping that they may be of interest to other teachers. I present only a rough outline that can be adapted according to local teaching and exam regulations and individual preferences.
My impression is that a course of this type can make students interested in existential risks and effective altruism for the longer term and perhaps shape the ideas of some students for their future careers. I do not yet have any data, however, to back this up.
Are there serious risks to the survival of human civilization and, if so, should we worry about them? To what extent should we fear, for example, supervolcano eruptions, asteroid impacts, natural and artificial pandemics, runaway climate change, nuclear war, unaligned superintelligence or perhaps even entirely new risks that are not yet on our radar?
In this course, we reflect on why the global "effective altruism" movement is so concerned about the long-term future in general and about existential risks in particular. But we also study concrete such risks, learn what the best available science says about them, discuss which of them are the most serious, and consider strategies for mitigating them. As it turns out, the most serious risks can only be understood by combining tools from the natural, social, and cultural sciences. In that sense, this is a truly interdisciplinary course.
The course is mostly based on the book
"The Precipice: Existential Risk and the Future of Humanity" by Toby Ord, Bloomsbury, 2020.
Further (partly optional) literature includes:
MacAskill, W., Effective altruism (unpublished, undated document, probably written for didactic purposes, not to be cited).
Yudkowsky, E., Cognitive biases potentially affecting judgment of global risks, in: Bostrom, N and Cirkovic, M. (eds.), Global Catastrophic Risks, Oxford University Press, 2008, pp. 91- 119.
The volume by Bostrom and Cirkovic from which Yudkowsky's essay is drawn is now slightly dated but still very useful. I recommend that students buy it but do not make it mandatory. I did the first iteration of this course before Toby's book appeared, based on that volume. I usually also recommend the books by Phil Torres and Bryan Walsh on existential risks, but I do not make reading them mandatory either.
Of course a large collection of articles on specific risks could be included here, for instance the articles on longtermism in this syllabus: https://forum.effectivealtruism.org/posts/NYxaxLZS5tiiyiHnG/a-full-syllabus-on-longtermism.
In this course I mostly use three class formats.
Class type I: Group exercises
Students work in breakout groups of 3-6 students on exercise sheets that I prepared for them beforehand. The questions on the exercise sheets are usually a mixture of questions that check whether the students properly understood the readings and questions that go beyond what can be answered based on the readings. For the latter questions, students either have to obtain additional information based on sources that they discover themselves, or they have to think independently based on the information provided in the readings. Students in each breakout group are supposed to agree on a joint answer to each question and to write it down in a concise manner. The teacher alternately supports the different groups in their work process.
At the end of class, after the group work session, I usually dissolve the breakout groups and gather all students in the same (physical or virtual) spot to discuss what they experienced as most remarkable while working on the exercise sheet.
Class type II: Group presentations
Students work in breakout groups on presentations covering specific aspects of this class's topic. I suggest suitable presentation topics beforehand and assign those topics to the different group numbers. Students can choose the topic (breakout group) that they wish to work on. This allows them to go a bit more in depth, contentwise, than the book does. Presentations are prepared in the first half and given to the rest of the overall group in the second half.
Class type III: (Guest) lectures
These are the only classes where the entire group of students stays together all the time. Lectures are given either by some qualified academic guest or by some student(s) as part of the overall assessment.
Note: How class types are assigned to sessions with specific topics in the schedule below is based on my experiences about what seemed natural and what seemed to work. Other assignments are possible and might be even better.
Class 1: Introduction
Student preparation: none (or: reading Ord, Ch. 1)
Activities:
1st half: teacher discusses the syllabus, outlines the schedule and assessment moments
2nd half: teacher and students introduce themselves and their backgrounds, voice their expectations in the course. Optional, to make introductions more personal: Participants may name the risk/worries that they are most afraid of and/or were most afraid of in their childhood or youth.
Class 2: Long-term trends and existential risks
Student preparation: reading Ord, Chs. 1 and 2
Activities:
Class type I, work in breakout groups on an exercise sheet using a shared document. The exercise sheet has questions on positive and negative long-term trends, "revolutions" in history, and longtermism. Questions are to be answered based on Ord, Chs. 1 and 2, and any other literature that the teacher might recommend and/or the students find useful.
Class 3: Effective altruism
Student preparation: Study some webpages related to effective altruism, e.g. 80000hours.org and its problem profiles
Activities:
Class type III, (guest) lecture on effective altruism, its different focus areas (near-term humans, animals, the long-term future), and the different criteria for identifying target activities (importance, tractability, negectedness). Q&A session.
Class 4: Psychological biases affecting our judgement about existential risks
Student preparation: Read Yudkowsky, try to identify cognitive biases encountered in everyday life or in the media during this week and make notes
Activities:
Class type I, work in breakout groups on an exercise sheet. The exercise sheet has questions on the cognitive biases encountered during the week and about properly understanding the text by Yudkowsky.
Class 5: Risks from nature
Student preparation: Read Ord, Ch. 3
Activities:
Class type II
1st half: Students prepare short presentations on three different natural risks in breakout groups.
Group 1: Asteroids and comets
Group 2: Supervolcanoes
Group 3: Stellar explosions
Each presentation should address the same points: what is the risk? How severe is it? By what causal mechanism is it a potentially serious risk? What are the energy scales involved? (How) Can the risk be mitigated?
2nd half: Students give the presentations prepared in the 1st half of class.
Class 6: Nuclear war
Student preparation: Read Ord, Ch. 4.1
Activities:
Class type II
1st half: Students prepare short presentations on different aspects of nuclear war in breakout groups.
Group 1: Physical foundations of nuclear weapons (mass-energy equivalence, fissile materials, chain reactions)
Group 2: Causes and consequences of nuclear winter
(Group 2a [optional]): Likely direct effects of nuclear detonations
Group 3: Treaties aimed at preventing the spread and use of nuclear weapons
2nd half: Students give the presentations prepared in the 1st half of class.
Class 7: Climate change
Student preparation: Read Ord, Ch. 4.2
Activities:
Class type I, students work in breakout groups on exercises related to climate change. The exercise sheet contains questions on the mechanisms behind climate change, the main (known and unknown) determinants of how dramatic climate change will be, and geoengineering.
(Optional class 7a: Class type III, guest lecture on climate change as a collective action problem, energy systems and energy system decarbonization. This is a topic that I have worked on myself and it is relevant when it comes to mitigating climate change, so I usually give a “guest lecture” here.)
Class 8: Environmental damage
Student preparation: Read Ord, Ch. 4.3
Activities: Class type II1st half: Students prepare short presentations on different aspects of environmental damage in breakout groups.
Group 1: The limits to growth hypothesis and ecological collapse: background, central assumptions, criticism
Group 2: Biodiversity loss: a serious threat to humans or merely an aesthetic issue -- or something in between?
Group 3: Freshwater scarcity: state of the art, future prospects, and the role of climate change
2nd half: Students give the presentations prepared in the 1st half of class.
Class 9: Natural and artificial pandemics
Student preparation: Read Ord, Ch. 5.1
Activities:
Class type I, students work in breakout groups on exercises related to pandemics. The exercise sheet contains questions on different types of pathogens, the history of pandemics, the features that a pathogen would have to have to threaten human extinction, laboratory outbreaks, and biological weapons. It can be helpful to use SARS-Cov2 as a "background case" with which everyone is familiar for reference and comparison.
Class 10: Dystopias
Student preparation: Read Ord, Ch. 5.2
Activities:
Class type I, students work in breakout groups on exercises related to dystopias. The exercise sheet contains questions on the distinction between different kinds of dystopias, the expected comparative likelihoods of those different types and about the possibility that attempts to combat existential risks might end up "locking in" some dystopian future. This class is also a good opportunity to broach the topic of "suffering risks" ("s-risks"), either by adding questions about those to the exercise sheet or including a student presentation.
Class 11: Unaligned artificial intelligence
Student preparation: Read Ord, Ch. 5.3
Activities:
I have experimented both with class type I (work on group exercises) and class type III (guest lecture) for this class. In both settings, the focus is on avoiding misunderstandings of what the suggested risks from unaligned AI really are, notably, that they are not about "killer robots" or AI suddenly turning sentient. It would be good to have another good introductory text here, beside the section from Toby Ord's book that links a bit more closely to the actual practice of AI research and applications, but I have not yet found one that is sufficiently short, dense, and accessible to qualify as the ideal mandatory additional reading. (Suggestions welcome!)
Class 12: Other risks and the risk landscape
Student preparation: Read Ord, Chs. 5.4 and 6
Activities:
Class type II, students form four breakout groups. In these groups they work on (i) potential risks from nanotechnology, (ii) possible risks from extraterrestrials, (iii) risks versus risk factors, and (iv) on how collective action problems drive catastrophic risks. Presentations are given and discussed in the second half of class.
Alternative for this class:
Plenary discussion in which the students together create a "geography" of the existential risks landscape -- including risk factors -- on the white board. This can help identify which items are particularly central to existential risk mitigation.
Class 13: Safeguarding humanity
Student preparation: Read Ord, Ch. 7
Activities:
Class type I, students work in groups on an exercise sheet. The exercise sheet contains questions about whether the most risky technologies tend to be inextricably linked to the most beneficial, whether stifling the development of a risky technology for a while can be possible or desirable, how the development of technologies can be slowed or accelerated, whether Ord's narrative concerning interplanetary, interstellar, and intergalactic civilization expansion is plausible, and what kinds of turns history might take that Ord does not foresee.
Class 14: Presentation session
Student preparation: Have presentations ready
This class can be structured like a mini-conference where students, in small groups or alone, give talks on topics related to existential risks that they have prepared in the weeks before. (See the section on assessment moments for more details.)
Class 15: Our potential -- final discussion
This session offers the opportunity of a more or less structured or free discussion of what to do in light of the work done in the previous classes. To stimulate the discussion, one can ask students such questions as: Having followed this course, do you find it plausible that combating existential risks is a worthwhile enterprise? Are you now more or less worried about existential risks than before following this course? What do you think might be the most effective measures that individuals/groups can take to mitigate existential risks? What are the most promising options for you personally, given your specific abilities?
If a judgement calibration test is used to check students' factual knowledge about matters learned in this course (see the section below on assessment moments), this session is a good occasion for that test.
Class 16: Essay/paper writing support session
For the final session of the course I usually ask students to submit an outline or draft version of their final essay or term paper beforehand. I provide feedback on what they submit, either before this session (via email) such that my feedback can be discussed in this session or directly during the session. The class can be arranged such that students who are not at that moment getting feedback by me provide peer feedback to each other. Having such a class usually dramatically improves the academic quality of students' work.
I have experimented and continue to experiment with different types of assessment moments in this course. Here are some suggestions.
Assessment moments to test students' knowledge/understanding of the subject matter of the course:
1. Multiple choice quizzes
In particular if one is worried that students may not prepare properly for classes it can be useful to signal from the very start that a thorough preparation for all classes is crucial. Scheduling a very short multiple choice test with questions that have a clear answer as provided in the readings for the respective class can be useful in this case. I usually try to strike a good balance in these questions between ones that concern relevant empirical information that students do not already know irrespective of whether they really did the readings and ones that concern key arguments that are made in the readings, to make sure that students really understand what the point of the considerations really is that they read. An overall grade can be awarded for this assessment moment that corresponds to the share of correct answers in the multiple choice tests.
2. Judgement calibration test
The Judgement Calibration test is supposed to do two things: first, make sure that students have really read the material and know its content; and second, test whether they can properly calibrate their confidence regarding the truth of their own answers.
Concretely, the test works like this: students receive a list of 30 (or more or fewer) statements. Each one is either determinately true or false, as can be known from having done the readings. For each statement, students specify their degree of belief (in percent) that the statement is true. The grade is computed according to the so-called Brier score -- the mean quadratic distance of their degree of belief from the actual truth value of the statement, expressed as 100% (for true statements) or 0% (for false statements). This enables students to make knowledge gaps comparatively harmless (they simply have to reply "50%"), but their grade will suffer, if they respond, say, 99% for a statement that is in fact false. This leads to a very different testing experience from an ordinary multiple choice test, one that actually fits better with various contexts where we must make decisions under uncertainty. It fits very well with the overall topic of the course.
So far, I have only made the plan for the judgement calibration test and never performed it in actuality. I have received permission from the University College Groningen's board of examiners, however, to use this assessment moment in the academic year 2021/22.
Other types of assessment moments are more useful to give students the opportunity to dive more deeply into specific topics around existential risks that they find particularly important, captivating, or relevant.
3. Group presentation
Groups of students (3-5 students per group) or individual students. If the group format is chosen, students should form the groups early on in the course. Presentation topics should go at least a bit beyond what is in the book, and students should get the opportunity to experience what it is like to delve into some aspects of existential risks research more deeply. Settling on a specific topic should be done after consultation with the teacher, making sure that the topic indeed connects well to the rest of the course and can realistically be introduced and addressed in that rather short period of time. Each group presentation should come with a short Q&A session. One can set up the presentation session a bit like a proper academic workshop, with a coffee break in between, perhaps joint lunch or dinner afterwards or later in the day.
4. Homework essay(s) and/or paper
Similar to the presentation, students can dive more deeply into one or several more specific aspects of existential risks here. Again, topic ideas should be approved by the teacher to make sure they are realistic and rewarding. Devoting one session of the course to teacher feedback and peer feedback (see Class 16 above) can help make sure that the quality of the papers ends up being high.
"2. Judgement calibration test
The Judgement Calibration test is supposed to do two things: first, make sure that students have really read the material and know its content; and second, test whether they can properly calibrate their confidence regarding the truth of their own answers."
This is really cool Simon, and awesome that you actually got permission to give actual grades by this mechanism. Curious how it works out in practice!
Great that you like this testing idea, Siebe!
I just tried out the judgment calibration test with my Munich students, with 25 relatively difficult-to-assess statements.
Main finding: Roughly half score better and half score worse than they would have done if they had just uniformly answered "50%". I guess that this indicates that the test was slightly too difficult. Notably, I had included many statements about orders of magnitude (e.g. energy released by the sun in a year, or time scales), and those seem challenging.
But the best students had a mean square deviation of the estimate from the truth value of about 16%, which I guess is quite good.