HPS 1616 — Artificial Intelligence and Philosophy of Science — Fall 2021

Instructor

Colin Allen, Professor, Dept. of History & Philosophy of Science, University of Pittsburgh <colin.allen@pitt.edu>

Course Description [Jump to Readings]

Artificial Intelligence (AI) is one of the core disciplines of cognitive science but with recent advances in machine learning (ML) it has become an increasingly pervasive, and some worry too invasive technology in society. The technology raises fascinating questions: Can machines think? Can reasoning be mechanized? Is AI really intelligence? Is ML really AI? What can we learn about the human mind from AI? Could machines be(come) conscious? Should AI be regulated, and how? We will survey some of the controversies and puzzles that AI has provoked.

Course Objectives

By the end of this course, you should have a grasp of the history of the field and the technologies underlying the current developments that have pushed AI (including ML) into the forefront of public discussion, as well as a framework for thinking about the scientific significance and the ethical and social implications of these technologies. No experience with computer programming is required for this course, but you will be expected to develop accurate, non-technical understanding of concepts such as algorithm, knowledge representation, machine learning, and artificial neural network. Given this knowledge you will be expected to reflect critically upon the claims about human nature and the future of humanity and society that are driven by the seemingly rapid developments in technology.

This course satisfies the University General Education requirement in PHILOSOPHICAL THINKING OR ETHICS.

Consultation Hours

Make use of open office hours! I enjoy talking about these issues, and you should not feel embarrassed if there's something you don't understand. There are no "stupid" questions, just questions and ideas that are sometimes hard to put into words, and that are best worked out in dialogue with others. Philosophy deals with hard questions and sometimes seemingly intractable problems. Philosophy of science further complicates things by being dependent on relatively sophisticated understanding of the relevant science.

Course Format, Assessments, and Attendance

We will follow all University guidelines for COVID mitigation (see Health & Safety Statement at end of syllabus), and we begin the semester in "classroom optional" mode. I intend to be in the classroom as much as possible. I will do my best to make sure those attending virtually can see and hear what happens in the classroom, but cannot be responsible for limitations in the technology. I will try to remember to record lectures, but don't always remember to start the recording. Feel free to remind me at the beginning of class.

Classes will be a mixture of lecture and discussion. I will sometimes use presentation slides for class, sometimes not — some material lends itself better to more structured presentation, some lends itself to a more free-flowing Q&A-driven classroom style. When I do use slides, they will be made available through the class Canvas site after class, but they will not reflect the full content of the lecture that day. So, you may use them for later study, but they are not a substitute for attendance, and taking your own notes if that's your style. Concerning note-taking and studying, you are advised to read the following two articles from the cognitive science of learning:

• Mueller & Oppenheimer 2014 The Pen is Mightier than the Keyboard
• Karpicke & Roediger 2006 Test-Enhanced Learning

You will be assessed on your ability both to accurately convey concepts covered in the readings and lectures in your own words, and to synthesize ideas from different areas into coherent arguments for and against claims involving AI. Your grade will be based on a writing portfolio that you will develop during the seminar. Although midterms and final will be scheduled, these will mainly serve as hard deadlines to do the writing that is asked of you, under timed conditions if you procrastinate to the end!

Your final grade for the course will be based on your full portfolio of work towards the final paper, which should be 10-12 pages (3500-4200 words) in one of these two formats:

1. A traditional philosophical essay (argumentative piece) that (a) describes a philosophical question or problem concerning AI, (b) outlines one or more positions that have been or could be reasonably taken on the question/problem, (c) critiques those positions, (d) and offers arguments for your own position (or indicates what further research is necessary to come to a position); or
2. A policy paper (similar to a "white paper" or "green paper") that (a) identifies a societal or ethical issue with AI, (b) surveys a range of approaches to dealing with the problem, (c) identifies pros and cons of each approach, and (d) makes a specific recommendation ("white paper") or provides a suite of acceptable options for policy makers to choose from ("green paper").

The main difference between these is that the first allows for more theoretical issues in philosophy of AI to be addressed whereas the second is a piece of applied philosophy designed to provide a practical solution to a current or imminent problem with AI.

In each case, you will build up the four parts on the schedule shown below, and you will have the opportunity to revise earlier parts at each submission checkpoint. Also, the checkpoints are designed so that you do not have to choose between theoretical and applied until the mid point of the semester. None of the stages will be formally graded, but detailed comments will be provided. Beginning at the midterm I will give informal indications of rough grade range (whole letter grades only) to which your work is trending. Course grade will be based on the final paper, meaning that improvement in response to feedback is fully taken into account. Each unauthorized earlier step that you skip (except the optional Nov 24 item) will result in an automatic deduction of one +/- grade level. So, in theory, you may turn in only the final paper and if it's perfect, receive a B- for the course. In practice, however, you should probably expect to do worse if you pursue this strategy.

From the midterm onwards you are expected to provide scholarly references. You may use any standard citation format (e.g., APA, Chicago) provided you are consistent. Most work will be submitted electronically through Canvas, although if you do any writing in class, it will be scanned and added to your Canvas portfolio.

1. Due Wed Sept 8
   Preliminary exercise: 1-2 paragraphs on what you currently don't know but would like to know about AI.
2. Due Wed Sept 22
   1-2 paragraphs: State as clearly as you can what theoretical or applied question about AI currently interests you the most, and what are the conceptual or theoretical issues that currently make answering this question difficult. Mention which course readings that you've encountered thus far are relevant to this interest, but you don't have to discuss them in any detail.
3. Due Wed Oct 13, 11:00 a.m. ("Midterm")
   2-3 pages: identify at least two responses to a clearly-stated problem in theoretical or applied philosophy of AI (i.e., provide a draft of parts (a) and (b) of your final report). You may choose a different problem than you submitted at step 2, or rework the previous one in light of feedback received. You should cite relevant sources for your disussion. If you have not submitted electronically via Canvas by 11:00 a.m., you are required to sit the exam in person under timed examination conditions.
4. Due Wed Nov 3 Mon Nov 8
   2-3 new pages for part (c) in which you critically evaluate available approaches to the problem previously outlined. If desired, you may combine with revisions of previous 2-3 pages corresponding to parts (a) and (b). You may also, if you prefer, completely switch topics without necessarily consulting me; just provide new material for all three parts, 4-6 pages. Cite sources thoroughly.
5. Optional by Wed Nov 24
   You may submit a draft of your section (d) for my feedback.
6. Due Friday Dec 17 10:00 a.m.
   Final paper of 10-12 pages. Include a properly formatted list of references (does not count towards page total). May be submitted by beginning of scheduled final time or written under timed conditions as scheduled by the university 10:00-11:50, location tba.

Missed Work Policy

Deadlines may be extended only in cases of predictable absences due to university-sanctioned activities or genuine emergencies.

• For predictable absences e.g., due to university activities such as student-athletic events, field trips in other classes, etc., or for religious observances, you should talk to me at least two weeks prior to the absence to make alternative arrangements.
• For emergencies such as illnesses, or deaths in the family, please let me know as soon as possible. In some circumstances you may be asked to provided written documentation.

Grade penalties for other missed assignemnts were explained above.

Missed Class Policy

Because you will be evaluated on your ability to synthesize ideas discussed in class and in the readings, you should not expect to do well if you do not attend lecture or do the readings, or if your writing fails to engage with ideas presented and discussed in class. However, this is not grade school, so attendance will not be officially enforced. In all cases of absence, excused or unexcused, it is your responsibility to get missed notes and information from a classmate.

Readings and Reading Schedule

You will need to obtain these books for this class:

1. Melanie Mitchell, 2019, Artificial Intelligence: a guide for thinking humans. Picador/Farrar, Straus and Giroux.
   [ Amazon link; Compare prices via campus bookstore.]
2. Wendell Wallach and Colin Allen, 2009, Moral Machines: Teaching Robots Right from Wrong MIT Press. [I have provided an electronic copy via Canvas for your personal use linked to the reading schedule below.. It is a copyright violation to distribute this copy elsewhere.]

Readings from these books will be supplemented with additional readings provided electronically through the links below. These items should be read in the order listed, but there is no set schedule. We will take the lectures and discussions at a pace that is necessary for mastery of the concepts and arguments. The →s in the list will be updated to show what you should be reading currently. As appropriate, too, I will insert extra readings to provide additional context and information, especially in the optional readings list.

Required Readings

There is no set schedule, but a (revisable) sequence as shown below. I will adjust the pace and add other readings according to class interests & understanding.
M# indicates chapters in the Mitchell book.
WA# indicates chapters in the Wallach & Allen book.
→ in the left column indicates everything you should have read by the date (before class time) in the second column. You are strongly encouraged, of course, to read ahead!

now	date	reading
→	09/01	M0 Prologue: Terrified
→	09/01	Superintelligence: Bostrom 1998
→	09/08	M1 The roots of AI
→	09/13	M2 Neural networks and the ascent of machine learning
→	09/13	M3 AI spring
→	09/13	Turing's Imitation Game(s): Sterrett 2000
→	09/20	M4 Who, what, when, where, why
→	09/20	"Deep learning: a philosophical introduction" Buckner 2019
→	09/22	M5 ConvNets and ImageNet
→	09/22	M6 A closer look at machines that learn
→	09/27	M7 On trustworthy and ethical AI
→	09/27	WA1-WA4 Moral Machines: Why?
→	10/04	"Gender Shades: Intersectional Accuracy Disparities in Commercial Gender Classification" Buolamwini & Gebru 2018
→	10/06	M8 Rewards for robots
→	10/06	M9 Game on
→	10/11	M10 Beyond games
→	10/18	M11 Words and the company they keep
→	10/20	M12 Translation as encoding and decoding
→	10/25	M13 Ask me anything
→	10/25	"On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?" Bender, Gebru, et al. 2021
→	11/01	M14 On understanding
→	11/03	M15 - Knowledge, abstraction and analogy in AI
→	11/08	"Explaining Explanations in AI" Mittelstadt et al. 2018
→	11/10	"Engineered Wisdom for Learning Machines" Karlan & Allen in prep.
→	11/15	WA5-WA8 Moral Machines: How?
→	11/29	WA9-WA12 Moral Machines: Beyond Reason
→	11/31	M16 Questions, answers and speculations

 

Optional readings for additional background [→ as before]

→ • Alan Turing 1950 Can a Machine Think?

→ • John Searle 1980 Minds, Brains, & Programs

→ • M. James 2017 What is a Turing Machine?

• G. Piccini and A. Scarantino (2011) “Information Processing, Computation, and Cognition” (with Andrea Scarantino), in Journal of Biological Physics, 37.1 pdf

• Newell & Simon 1975 Computer Science as Empirical Enquiry

• Schank & Abelson 1977 Scripts, Plans and Goal Understanding

• Ferrucci et al. 2010 Building Watson

• Doug Hofstadter on why Watson is not real AI and a counterpoint from two AI researchers

Blogs and other media on GPT

For an extensive list of technical papers on AI ethics/fairness issues see Prof. Dan Jurafsky's Stanford Course reading list. [h/t to Rohit Bommisetti]

Everything below is not specific to this course, but required/recommended for all Pitt syllabi.

Students in this course will be expected to comply with the University of Pittsburgh’s Policy on Academic Integrity. Any student suspected of violating this obligation for any reason during the semester will be required to participate in the procedural process as outlined in the University Guidelines on Academic Integrity. When you submit assignments with your name on them in this course, you are signifying that the work contained therein is all yours, unless otherwise cited or referenced. Any ideas or materials taken from another source for either written or oral use must be fully acknowledged. If you are unsure about the expectations for completing an assignment or taking a test or exam, be sure to seek clarification beforehand.

To learn more about Academic Integrity, visit the Academic Integrity Guide for an overview of the topic. For hands-on practice, complete the tutorial on Understanding and Avoiding Plagiarism.

Diversity and Inclusion

The Americans with Disabilities Act (ADA) is a federal anti-discrimination statute that provides comprehensive civil rights protection for persons with disabilities. Among other things, this legislation requires that all students with disabilities be guaranteed a learning environment that provides for reasonable accommodation of their disabilities. If you have a disability for which you are or may be requesting an accommodation, you are encouraged to contact both your instructor and Disability Resources and Services (DRS), 140 William Pitt Union, 412-648-7890, drsrecep@pitt.edu, 412-228-5347 for P3 ASL users, as early as possible in the term. DRS will verify your disability and determine reasonable accommodations for this course.

The University of Pittsburgh does not tolerate any form of discrimination, harassment, or retaliation based on disability, race, color, religion, national origin, ancestry, genetic information, marital status, familial status, sex, age, sexual orientation, veteran status or gender identity or other factors as stated in the University’s Title IX policy. The University is committed to taking prompt action to end a hostile environment that interferes with the University’s mission. For more information about policies, procedures, and practices, see Pitt's Civil Rights & Title IX Compliance pages.

I ask that everyone in the class strive to help ensure that other members of this class can learn in a supportive and respectful environment. If there are instances of the aforementioned issues, you may contact the Title IX Coordinator, by calling 412-648-7860, or e-mailing titleixcoordinator@pitt.edu. Reports can also be filed online: https://www.diversity.pitt.edu/make-report/report-form. You may also choose to report this to a faculty/staff member; they may also be required to communicate about such issues to the University’s Office of Diversity and Inclusion. If you wish to maintain complete confidentiality, you may also contact the University Counseling Center 412-648-7930.

Statement on Classroom Recording

To ensure the free and open discussion of ideas, students may not record classroom lectures, discussion and/or activities without the advance written permission of the instructor, and any such recording properly approved in advance can be used solely for the student’s own private use.

At certain times, lectures or portions of the lectures may be recorded by the instructor. Before starting recording, it will be announced to the class. Students who do not wish to be identifiable during such recordings may remain silent and obscure their faces either by turning off their own video feed if connected via Zoom or obscuring their faces if in the classroom.

Copyright Notice

Materials provided for the course may be protected by copyright. United States copyright law, 17 USC section 101, et seq., in addition to University policy and procedures, prohibit unauthorized duplication or retransmission of course materials. See Library of Congress Copyright Office and the University Copyright Policy.

During this pandemic, it is extremely important that you abide by the public health regulations, the University of Pittsburgh’s health standards and guidelines, and Pitt’s Health Rules. These rules have been developed to protect the health and safety of all of us. Universal face covering is required in all classrooms and in every building on campus, without exceptions, regardless of vaccination status. This means you must wear a face covering that properly covers your nose and mouth when you are in the classroom. If you do not comply, you will be asked to leave class. It is your responsibility have the required face covering when entering a university building or classroom. For the most up-to-date information and guidance, please visit coronavirus.pitt.edu and check your Pitt email for updates before each class.

If you are required to isolate or quarantine, become sick, or are unable to come to class, contact me as soon as possible to discuss arrangements.