Philosophy of Computer Science:
What I Think It Is, What I Teach, & How I Teach It
William J. Rapaport
Department of Computer Science & Engineering,
Department of Philosophy,
and Center for Cognitive Science
[email protected]
http://www.cse.buffalo.edu/~rapaport
Outline
•
•
•
•
What is the Philosophy of Computer Science?
What should be taught in a PhilCS course?
How can it be taught to non-philosophers?
How can a writing-intensive course be taught
in a lecture environment?
What Is “Philosophy of CS”?
• xy[ y = philosophy of x
= philosophical investigation of:
–
− the nature of x,
− x’s assumptions,
− x’s methods,
− x’s goals, etc.]
Let x = CS / Philosophy of CS exists!
• Philosophy of CS : CS = Phil of Science : Science
= Phil of Physics : Physics
etc.
• Philosophy of CS ≠ Philosophy of AI
– Philosophy of AI  Philosophy of CS
• Philosophy of CS ≠ Philosophy of Information
– Machlup & Mansfield 1983; Floridi 2002
– Philosophy of CS  Philosophy of Info ≠ 
Is Anyone Doing PhilCS?
•  lots of philosophical essays on the topics
covered by PhilCS
– But few authors call it ‘PhilCS’
•  some courses in PhilCS
– But not many (as of ~2005):
• Swedish National Course on PhilCS
– Gordana Dodig-Crnkovic @ Mälardalen U.
• Topics in PhilCS
– Eli Dresner @ Tel Aviv U.
• Phil of Computing
– Bernard W. Kobes @ Arizona State U.
• ∃ lots of “PhilCS” courses that are really PhilAI
So, What Is the Philosophy of CS?
(& How Can It Be Taught?)
0.
1.
2.
3.
6.
7.
What is philosophy?
What is computer science (or computing science)?
What is a computer?
(the “hardware question”)
What is computation/computing?
What is an algorithm? What is a computer program?
(the “software question”)
What is the relationship between HW & SW?
(the “mind/body question”?)
What is the relation of computers/computation
to the real world?
Philosophy of Artificial Intelligence
Computer Ethics
8.
Etc.?
4.
5.
•
Exercise for the audience: Name some other topics!
What Is Philosophy?
• Necessary topic in a CS course
– Cf. Perry’s Scheme of Intellectual and Ethical Development
• Dualism
– Correct answers to all questions are known only to Authorities
 Multiplism
•  answers / all opinions are equally good
 Contextual Relativism
• Validity of an opinion is relative to its evidential context
• Are CS students “Dualists”?
• See (fear?) philosophy as being “Multiplistic”
• Is philosophy a “Contextually Relativistic” discipline?
 Students must critically evaluate opinions on basis of evidence
What Is Philosophy? (cont’d)
• Philosophy =def search for truth in any field
by rational means
– “rational” =
• (deductive) logic
• empirical, scientific investigation
• Philosophy of X = study of fundamental
assumptions, methods, goals of X
• Socrates/Plato:
– philosopher as gadfly who challenges assumptions
What Is Philosophy? (cont’d)
• Critical thinking & informal argument analysis
– Computing Curricula 2001
• “Social & Professional Issues” knowledge area
– “Methods & tools of analysis”
• Readings on nature of philosophy:
–
–
–
–
Audi, “Brief Guide for Undergrads” (2001)
Plato, Apology
Colburn 2000, Chs. 3–4 (on history)
Woodhouse, Preface to Philosophy (2003)
• < 1 class period!
What Is Computer Science?
• ∃ philosophical & political motivations
• Philosophy: What is the “nature” of CS?
– Probably, CS is not a natural kind
 CS = what computer scientists do!
• Extensional characterization
– But: can still give intensional version of this extension:
• What do computer scientists do? (see below)
• Academic politics:
– Which dean should oversee CS?
• (Arts &) science?
• Engineering?
• CS/informatics?
What Is CS? (cont’d)
• Science of computers & surrounding phenomena (including algorithms)
– Newell, Perlis, & Simon 1967
• Study of algorithms & surrounding phenomena (including computers)
– Knuth 1974
• What can be automated/solved by TM/expressed recursively?
– AI = Is cognition recursive?
– Arden 1983
• Artificial science (empirical study, but not a “natural” science)
of phenomena surrounding computers
– [Newell &] Simon 1976, 1969/1996
• Natural science of procedures
– Shapiro 2001
[McCarthy 2006: computational procedures]
• Body of knowledge dealing with info-transforming processes
– Denning et al. 1989
• Study of information
– Hartmanis & Lin 1992
• Engineering (not science)
– Brooks 1996, Loui 1987
• Study of virtual phenomena
– Crowcroft 2005
Is CS Science or Engineering?
•
Motivation:
Local (UB) politics
1. What is science?
– Purpose of science:
•
•
To describe the world?
To explain it?
– Are scientific theories…
•
•
Instrumental (i.e., “calculational”)?
Realistic?
– Is the scientific method…
•
•
Experimental & cumulative?
Driven by paradigms & revolutions?
– Is mathematics a science?
– Readings:
•
•
Papineau, “Philosophy of Science” (1996)
Kemeny, A Philosopher Looks at Science (1959)
Is CS Science or Engineering? (cont’d)
2. What is engineering?
•
Application of science to technology?
– invention of devices vs. discovery of new knowledge?
•
Defined by a professional education?
– Davis, Thinking Like an Engineer (1998)
•
A design activity?
–
•
Petroski, “Early [Engineering] Education” (2003)
Is CS a new kind of engineering?
–
Studies theory, design, analysis, & implementation of
information-processing algorithms
•
Loui, “CS Is an Engineering Discipline” (1987)
If CS Is a Science, then…
• What is it a science of?
– Of computers?
• What is a computer?
• Cf. microscopy
– a discipline that no longer exists!
– Of computation?
• What is computation?
What Is a Computer?
(Historical Perspective)
• Survey of history of computers (& computation)
– 2 parallel goals:
• To make calculation easier/mechanical—to automate it
– Pascal, Leibniz, Babbage, Turing, Atanasoff & Berry,
Eckert & Mauchly, von Neumann, …
» Asprey, Computing before Computers (1990)
• To provide a foundation for mathematics
– Leibniz, Boole, Frege, Hilbert, Gödel, Turing, …
» Davis, Engines of Logic (2000)
What Is Computation / What Is an Algorithm?
(Mathematical Perspective)
•
Function viewed extensionally:
–
•
Function f is computable ≈def
–
–
•
Set of input-output pairs (s.t. same I/P yields same O/P)
 “algorithm” A that computes f
i.e.,
( i )[ A(i) = f(i) ]
& A specifies how i and f(i) are related
An algorithm A for a problem P ≈def
–
finite procedure (= set of instructions) for solving P that is:
1. unambiguous
2. halts
3. outputs correct answer to P
•
“Slow reading” of Turing 1936
–
Turing’s & Church’s Theses
•
TM ≡ -definability (≡ etc.)
What Is CS? (revisited)
• Possible answer:
– CS = study of algorithms
and computers that implement them
– Link between these 2 “branches” of CS
(i.e., where the two histories intersect):
• Turing’s 2 computers (TM, ACE)
• ACE was an implementation of TM
– With implementation-dependent details
– With implementation-dependent limitations
3 Great Insights of CS
Boole’s & Shannon’s Insight
1.
•
Only 2 nouns are needed to represent “anything”:
“0”, “1”
Turing’s Insight
2.
•
Only 5 verbs are needed to manipulate them:
i.
ii.
iii.
iv.
v.
Move-left
Move-right
Print-0
Print-1
Erase
Boehm & Jacopini’s Insight
3.
•
Only 3 rules of grammar are needed to combine these:
i. Sequence
ii. Selection
iii. Repetition
Also useful &/or elegant:
Exit
Named procedures
Recursion
What Is a Computer?
(Philosophical Perspective)
• Searle, “Is the Brain a Digital Computer?” (1990)
– Everything is (can be seen as) a digital computer
– What counts is how they are used
• Hayes, “What Is a Computer?” (1997)
– Not everything is a computer
– A computer is (like) “magic paper”:
• Input = patterns describing changes to (other) patterns
• Output = the results of making the changes
• Cf.: a device that changes assignments to variables (Thomason 2003)
• Is the universe a computer?
– Does the solar system compute Kepler’s laws?
– Lloyd & Ng, “Black Hole Computers” (2004)
What Is an Algorithm?
(Philosophical Perspective)
• What is a procedure?
• Algorithms ≠ recipes
• But recipes ≈ specifications (Preston, unpublished)
– Different implementers (e.g., chefs) fill in details differently
– Allow for improvisation (cf. jazz, rock)
• “Mundane”/“quotidian” procedures (Cleland 1993ff):
– Effective procedures that generate causal processes
• E.g., recipes
– But not TM-computable
• Because effectiveness depends on external world
Are There Other Kinds of Computation?
• Computation of functions that are not TM-computable
– Copeland, “Hypercomputation” (2002)
• Turing:
oracle machines
– Add external source of non–TM-computable information
(“6th verb”: non–TM-computable basic operation)
• Boolos & Jeffrey: “Zeus” machines
– Infinitely accelerating
• Bringsjord 1995ff:  computationalism is doomed
• Wegner:
“interaction” machines
– Non-halting procedures with I/P from external world
– E.g., ATM, airline-reservation systems
• Putnam, Gold:
“trial & error”/inductive-inference machines
– Like TM, but last answer counts, not first
– May be needed for AI to succeed
• Kugel, “Computing Machines Can’t Be Intelligent (& Turing Said So)” (2002)
What Is a Computer Program?
•
•
•
•
•
What is implementation?
Are some programs (scientific) theories?
What is software vs. hardware?
Can software be patented? Copyrighted?
Can programs be verified?
What Is Implementation?
• Ubiquitous notion, rarely defined
– Examples:
•
•
•
•
Programs “implement” algorithms
ML programs “implement” programs in high-level languages
Data structures “implement” ADTs
ADTs “implement” other ADTs
– Related to “realization” (in philosophy of mind)
• Is implementation…
– A relation between…
• An “abstraction” & something “concrete”?
• 2 “abstractions”?
– An isomorphism? A homomorphism?
• Readings:
– Chalmers, “On Implementing a Computation” (1994)
– Rapaport, “Implementation Is Semantic Interpretation” (1999, 2006)
• Syntax, semantics, formal systems (see below)
What Is the Relation of a Program to What It Models or Simulates?
• Can a program be a (scientific) theory?
– Pylyshyn, Johnson-Laird, Newell & Simon:
• Cognitive theories best expressed as computer programs
– In addition to statistical, mathematical, or natural languages
– Programs are simultaneously theory & model (implementation)
– Theory can be tested by executing program
• Philosophy of AI question:
–
Do such cognitive programs
• actually exhibit (i.e., implement)
• or merely simulate
cognitive processes?
• Readings:
– Weizenbaum, Computer Power & Human Reason (Chs.5,6) (1976)
– Simon, Sciences of the Artificial (Ch.1) (1996)
• Local UB color:
– Is a program that can identify handwriting a scientific theory of handwriting?
– Should its programmer be an expert witness on handwriting?
What Is Software?
• Software is a computer program changeable by a person
 Changeable hardwiring is software
– Moor, “3 Myths of CS” (1978)
• Software is syntactic form (see above)
– Suber, “What Is Software?” (1988)
• Software is a “concrete abstraction”:
– Software has both:
• Medium of description: text (abstraction)
• Medium of execution: implemented in electronics (concrete)
– Colburn, “Software, Abstraction, Ontology” (1999/2000)
Can/Should Software/Hardware
Be Patented/Copyrighted?
• Combines legal, social, ontological issues!
– Could be unifying theme for course
• If computer program is text, then copyrightable
– But not patentable!
– Yet exportable.
• “Same” program engraved on CD-ROM ( executable)
is a machine
 Patentable
– But not copyrightable or exportable!
• ∃ mismatch; something’s got to give
• Readings:
– Newell, “The Models Are Broken” (1985/86):
• CS needs better ontological theories of computational entities
– Koepsell, Ontology of Cyberspace (2000):
• Lawyers need to devise better methods of legal protection
Can Programs Be Verified?
• Background:  formal methods for proving program “correctness”
– Gries, Science of Computing (1981)
– Dijkstra, “Guarded Commands…” (1975)
• Smith, “Limits of Correctness” (1985)
– Should be required reading!!!
– ∃ gap between world & our models of it
• Can’t talk about real world except via a model or theory
– Computers are doubly removed from real world:
• Rely on models of models
• Yet must act in real world
• Fetzer, “Program Verification: The Very Idea” (1988)
– At best, can verify algorithms, not programs
– Can’t logically prove that causal systems won’t fail
– NB: requires firm grasp of “algorithm/program/implementation”
Philosophy of AI: Could We Build “AI”s?
• Deserves course of its own!
– My own AOS; lots of student interest
– Small fraction of the course (~ 1 week)
• What is AI?
• What is the relation of computation to cognition?
– Turing Test (Turing 1950)
• Computers will be said to be able to think if we can’t distinguish their
linguistic/cognitive ability from a human’s
• Arguably, PhilCS = Turing 1936 + 1950 :-)
– Chinese-Room Argument (Searle 1980)
• Computer could pass a TT without really being able to think
– Local color: “Syntactic Semantics” (Rapaport 1985ff):
• Syntactic symbol manipulation (what computers do well)
suffices for semantic interpretation
of the kind needed for computational cognition
Computer Ethics
•
Deserves a course of its own!!
–
•
Small fraction of the course (~ 1 week)
Moor, “What Is Computer Ethics?” (1985):
–
–
Need metaphysical/ontological theories of computers…
in order to answer ethical/social questions
about their nature & use
Computer Ethics (cont’d)
1. Should we trust decisions made by computers?
•
Moor, “Are There Decisions Computers Should Never Make?”
(1979)
–
–
•
Friedman & Kahn,
“People Are Responsible; Computers Are Not” (1997):
–
•
No, as long as their track record is better than humans’
Up to us humans to accept/reject computer recommendations
Only humans can be moral agents
Johnson, “To Err Is Human” (2002)
–
Case where computer’s correct decision was overridden by human
Computer Ethics (cont’d)
2. Should we build “intelligent” computers?
•
Lem, “Non Serviam” (1971)
– What might happen if you create ALife
(and then lose your funding)
•
LaChat, “AI & Ethics” (1986)
– Perhaps we shouldn’t
– But considering the possibility allows us to deal with:



What is a person?
Should an AI with personhood have rights?
Could it be moral?
Summary & Unifying Theme
• 2 overview articles:
– Scheutz, “Philosophical Issues about Computation” (2002)
– Smith, “Foundations of Computing” (2002)
• Unifying theme:
– Relation of abstract computation to real world
•
•
•
•
•
•
Cleland’s “mundane procedures” causally affect real world
Smith on limits of computation
Fetzer on program verification
Implementation
Software vs. hardware
Copyright vs. patent
Where & Why Teach PhilCS?
• Philosophy department?
• Or: Computer Science department?
• Yes!
– Good intro to CS issues for philosophy students
• Useful way to bring wide variety of topics together…
• & can shed new light on classical philosophical problems in:
– Metaphysics & ontology
– Epistemology
– Ethics
• ∃ topics that are unique to (Phil)CS
– Good intro to philosophy for CS students
• Capstone course for senior-level CS students
• Overview course for entry-level CS students
Texts, etc.
• Floridi, Philosophy & Computing (1999)
• Colburn, Philosophy & Computer Science (2000)
– Both are monographs / Not neutral / Not ideal for intro course (?)
– Small  with my topics
• Floridi, Blackwell Guide to Philosophy of Computing and Information (2004)
– Anthology of original overviews (not classical papers)
– Small  with my topics
• Etc.:
– Monist 82(1) (1999) on PhilCS
– Minds & Machines, JETAI, CAP conference proceedings
– Websites:
•
•
•
•
Taylor, “Computational Philosophy”
Floridi’s homepage
Eden & Turner, “Philosophy of Computer Science”
My course :-)
How to Teach Philosophy of CS
• Thinking is best done by:
– Slow & active reading
– Discussion
– Writing (lots of it!)
• Expected 10-15 students, seminar setting
–
–
–
–
30 pre-registered (gut course?)
So, posted news about lots of writing
Enrollment increased to 60!
Settled at ~50!!
• No TA to help grade
• No recitation sections
• (Not a unique problem in 4-year colleges!)
• A problem, nevertheless!
Solution
(inspired by Lewis White Beck)
I.
Required:
•
Attendance + reading journal
–
II.
Max course grade = C
Optional:
•
5 short position papers
–
Max course grade = B
III. Extra-optional assignment
•
Term paper XOR final exam
–
Max course grade = A
Solution (I):
Required Attendance & Reading Journal
1. Attend & participate in all class discussions
(including “peer editing”)
2. Maintain a “reading journal”:
–  reading assignment:
• copy interesting passages
• comment on them
– To enforce (& substantiate) “active reading”
• Including writing and thinking
Solution (II): Optional Short Position Papers
• 5 1-page position papers
– Every 2-3 weeks
– 1 week to write first draft
• Due-date = “peer editing” day
– Students bring 5 copies of paper
– Small-group discussions
• 1 week later: revision due
• I graded ~40% of these & recorded the rest
– Each student got 2 papers fully critiqued & graded
– Could re-revise for higher grade
Position Paper #1: What Is CS?
•
Dean of Engineering says:
–
CS department should be moved
from Science to Engineering because:
1. Science = systematic observation, description,
experimental investigation, & theoretical
explanation of natural phenomena.
2. CS is the study of computers & related phenomena.
3.  CS is not a science
•
How would you respond to the Dean of
Engineering’s argument?
How to Evaluate an Argument
• Valid?
– Missing premises?
• Agree with premises (including missing ones)?
– What is science?
– What is CS?
• Give reasons for opinions!
Peer-Editing Instructions
1. Form small groups; share papers.
2. For each paper (10–15 minutes each), do:
a) Imagine peers as members of committee to make
recommendation to Provost
i. Goal = to help author clarify beliefs & arguments…
ii. …so that recommendation made on purely logical grounds
b) Author states/reads beliefs and reasons
c) Peers ask questions:
i. Why did you say ___ rather than ---?
ii. What did you mean when you said ___? [Etc.]
d) Don’t get defensive; peers are critical, but friendly.
e) Keep written record of questions/replies/advice
3. At home, revise paper; revision due in 1 week.
Advantages of Peer Editing
• Lots of opportunity for discussion among students
• Opportunity for me to interact with students 1-1
– I roamed & facilitated
• Multiple feedback on papers
– From peers as well as “Authority”
• Opportunity for critical thinking
– Paper topics = evaluation of an argument
• I only saw second draft!
Grading Position Papers
• Student observation:
– Argument-analysis format easier to grade than “ordinary” essay
• because:
– Actually:
fewer degrees of freedom
students saw role of evaluation more clearly
• Triage Theory of Grading (inspired by Paul Vincent Spade)
– Essays are either:
• clearly acceptable
• clearly lousy
• somewhere in between
A
F
C
Grading Position Papers
• Student observation:
– argument-analysis format easier to grade than “ordinary” essay
• because:
– Actually:
fewer degrees of freedom
students saw role of evaluation more clearly
• Triage Theory of Grading (inspired by Paul Vincent Spade)
– Essays are either:
•
•
•
•
clearly acceptable
clearly lousy
somewhere in between
[ not done
A
F [ → D]
C
F]
Grading Position Papers (cont’d)
The Triage Theory (cont’d)
• Any part of an assignment is either:
–
–
–
–
clearly acceptable
in between
clearly lousy
not done
A
C
D
F
(full credit)
(partial credit)
(minimum credit)
(no credit)
• Grade in “quantum units”
– Ends requests for “one extra point”
• Applicable to argument-analysis papers:
– valid? / why?
– missing premises?
–  premise, true?/agree?
• give argument for/against
A..F
A..F
A..F
– Then compute cumulative grade
• Weighted sum or average
Solution (III):
Extra Optional Assignments
• Term paper
– Topic approved in advance
– Default topics:
• encyclopedia article on PhilCS (for “Dualists”)
• your own reasoned answers to syllabus questions
(for “Multiplists”)
XOR
• Take-home, short-answer, essay-style final exam
– Analytical & evaluative summary
of possible answers to syllabus questions
Statistics
98%:
position papers
>80%:
TP xor FE
~70%:
FE (!)
Midsemester Course Evaluation
(inspired by Stuart C. Shapiro)
•
Midsemester more useful than end of semester
–
•
•
Should be followed by course correction
2 questions
(or 4 questions):
1.
2.
What aspects of the course would you like to see changed?
What aspects of the course do you especially like?
3.
& 4. Ditto for recitation section
Summarized & discussed/posted responses
Likes & Dislikes
• Chief complaint:
– Not enough time to do all the very interesting
reading!
• At midsemester, changed to:
1(or 2) required readings per topic or class
+ 1 strongly recommended
+ 1 recommended
Likes & Dislikes
• Likes:
– position papers:
• writing
• peer editing
• revising
– discussions
– skill & practice in
critical analysis/informal-argument evaluation
– some students continued use of reading-journals
afterwards
– website
I’d like to thank you for putting together such a great
course this semester….I never had much respect for
philosophy in the past—but this course has provided
me with an entirely new perspective….I learned as
much in your course as any other I’ve taken in my
graduate career at UB (not to mention the fact that
the skills I learned in [it] are far more transferable
than the skills of the more esoteric CS courses)….
I urge [you] to offer this course again….It offers
exactly the kind of breadth of education that the
department needs to stress, and with its CS flavor,
it can tap the interest of students….Please consider
making Philosophy of CS a regular offering :)
Conclusion
• PhilCS is a legitimate branch of philosophy
• Worth teaching to both CS & philosophy students
– CS students get to:
• think about new issues
or issues not discussed elsewhere in CS curriculum
• think critically
• find out what philosophy is like
– Philosophy students get opportunity:
• to learn about computers, computing, and CS
• to apply philosophical skills & knowledge to a relatively new
domain.
Reference & Website
• Rapaport, William J. (2005), “Philosophy of
Computer Science: An Introductory Course”,
Teaching Philosophy 28(4): 319–341.
• http://www.cse.buffalo.edu/~rapaport/Papers/philcs-complete.pdf
– contains lots of material not in published version, from course website
• Course website:
– http://www.cse.buffalo.edu/~rapaport/philcs.html
– (or Google “philosophy of computer science” :-)
Descargar

philcs-nacap - University at Buffalo