Cookbook science vs. explanatory science
Examples
Reasons
Problems
Is Computer Science Becoming a Cookbook
Science?
Bengt Nordström
Computing Science, Chalmers and University of Göteborg
Ämneskonferens i datavetenskap och numerisk analys i
Stockholm
12-14 juni 2006
Alternative?
Cookbook science vs. explanatory science
Examples
Reasons
Cookbook Science
Cookbook:
Study a number of cooks.
Summarize what they do by writing rules.
Problems
Alternative?
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Cookbook Science
Cookbook:
Study a number of cooks.
Summarize what they do by writing rules.
Cookbook Science:
Study how things work in practice (how nature is).
Summarize by writing rules.
Alternative?
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Alternative?
Carried over to Computing Science:
Study how things work in practice:
study existing programmers how they program.
study existing programming languages (what properties they
have)
study existing ways of structuring big programs (program
libraries in C or Java).
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Explanatory Science
The alternative to Cookbook Science is Explanatory Science:
You not only summarize by rules,
but also try to explain the causes.
Alternative?
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Alternative?
Explanatory Science
The alternative to Cookbook Science is Explanatory Science:
You not only summarize by rules,
but also try to explain the causes.
A good cook knows not only that you put just some drops of oil in
the egg yolk, but also why!
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Alternative?
Explanatory Science
The alternative to Cookbook Science is Explanatory Science:
You not only summarize by rules,
but also try to explain the causes.
A good cook knows not only that you put just some drops of oil in
the egg yolk, but also why!
Poincaré:
Knowledge is the end and action is the means.
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Computer Science as a science
First observation:
Computing Science is not a science in the sense of natural
science.
Alternative?
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Computer Science as a science
First observation:
Computing Science is not a science in the sense of natural
science.
We do not study Nature.
Alternative?
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Alternative?
Computer Science as a science
First observation:
Computing Science is not a science in the sense of natural
science.
We do not study Nature.
We are very little interested in physical, chemical, biological,
geological properties.
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Alternative?
Computer Science as a science
First observation:
Computing Science is not a science in the sense of natural
science.
We do not study Nature.
We are very little interested in physical, chemical, biological,
geological properties.
The material reality which we are interested in has very simple
properties.
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Alternative?
Computer Science as a science
First observation:
Computing Science is not a science in the sense of natural
science.
We do not study Nature.
We are very little interested in physical, chemical, biological,
geological properties.
The material reality which we are interested in has very simple
properties.
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Alternative?
Computer Science as a science
First observation:
Computing Science is not a science in the sense of natural
science.
We do not study Nature.
We are very little interested in physical, chemical, biological,
geological properties.
The material reality which we are interested in has very simple
properties.
The objects of our study are creations of human beings. In natural
science we study Nature, it is there and we want to understand
more and more about it. But in Computing Science we also create
our objects of study.
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Alternative?
Computer Science as a science
First observation:
Computing Science is not a science in the sense of natural
science.
We do not study Nature.
We are very little interested in physical, chemical, biological,
geological properties.
The material reality which we are interested in has very simple
properties.
The objects of our study are creations of human beings. In natural
science we study Nature, it is there and we want to understand
more and more about it. But in Computing Science we also create
our objects of study.
All man made objects are not worthy of study.
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Things taken for given (as part of Nature?):
Programming languages (C, Java, ML, Haskell, . . . )
Alternative?
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Things taken for given (as part of Nature?):
Programming languages (C, Java, ML, Haskell, . . . )
Specification languages (UML, VHDL, . . . )
Alternative?
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Things taken for given (as part of Nature?):
Programming languages (C, Java, ML, Haskell, . . . )
Specification languages (UML, VHDL, . . . )
Data description languages (XML, . . . )
Alternative?
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Things taken for given (as part of Nature?):
Programming languages (C, Java, ML, Haskell, . . . )
Specification languages (UML, VHDL, . . . )
Data description languages (XML, . . . )
Be friend with your bugs!
Alternative?
Cookbook science vs. explanatory science
Examples
Reasons
Programming language example: C, Java
Problems
Alternative?
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Alternative?
Programming language example: C, Java
C is a good assembly language, and Java is widely used. These
languages have caused a lot of practical problems, some of them
are scientific interesting, but many of them are just problems from
bad design. Two fields of Computing where this causes problems:
analysis of programming languages (semantics, static analysis)
program verification
Cookbook science vs. explanatory science
Examples
Reasons
Programming language example: Haskell
Problems
Alternative?
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Programming language example: Haskell
Official propaganda:
You reason about Haskell programs exactly like you reason in
mathematics, you can replace equals by equals etc.
Alternative?
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Programming language example: Haskell
Official propaganda:
You reason about Haskell programs exactly like you reason in
mathematics, you can replace equals by equals etc.
The truth:
Alternative?
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Programming language example: Haskell
Official propaganda:
You reason about Haskell programs exactly like you reason in
mathematics, you can replace equals by equals etc.
The truth:
This is simply not the case. Examples:
The seq construct.
Alternative?
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Alternative?
Programming language example: Haskell
Official propaganda:
You reason about Haskell programs exactly like you reason in
mathematics, you can replace equals by equals etc.
The truth:
This is simply not the case. Examples:
The seq construct.
Definition by pattern matching. (Well typed programs can go
wrong. Bad conversion rule)
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Alternative?
Programming language example: Haskell
Official propaganda:
You reason about Haskell programs exactly like you reason in
mathematics, you can replace equals by equals etc.
The truth:
This is simply not the case. Examples:
The seq construct.
Definition by pattern matching. (Well typed programs can go
wrong. Bad conversion rule)
Confusion between coinductive and inductive data.
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Alternative?
Programming language example: Haskell
Official propaganda:
You reason about Haskell programs exactly like you reason in
mathematics, you can replace equals by equals etc.
The truth:
This is simply not the case. Examples:
The seq construct.
Definition by pattern matching. (Well typed programs can go
wrong. Bad conversion rule)
Confusion between coinductive and inductive data.
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Alternative?
Programming language example: Haskell
Official propaganda:
You reason about Haskell programs exactly like you reason in
mathematics, you can replace equals by equals etc.
The truth:
This is simply not the case. Examples:
The seq construct.
Definition by pattern matching. (Well typed programs can go
wrong. Bad conversion rule)
Confusion between coinductive and inductive data.
There is no formal semantics of Haskell.
Cookbook science vs. explanatory science
Examples
Reasons
Specification language example: UML
Problems
Alternative?
Cookbook science vs. explanatory science
Examples
Reasons
Specification language example: UML
First standardize something fuzzy.
Problems
Alternative?
Cookbook science vs. explanatory science
Examples
Reasons
Specification language example: UML
First standardize something fuzzy.
Then impose its use in industry.
Problems
Alternative?
Cookbook science vs. explanatory science
Examples
Reasons
Specification language example: UML
First standardize something fuzzy.
Then impose its use in industry.
Then try to give semantics to it.
Problems
Alternative?
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Specification language example: UML
First standardize something fuzzy.
Then impose its use in industry.
Then try to give semantics to it.
General problem with specification language independent of
programming language.
Alternative?
Cookbook science vs. explanatory science
Examples
Reasons
Data description language example: XML
Problems
Alternative?
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Data description language example: XML
Some problems:
Unorthogonal design. Strange restrictions based on poor
understanding or inefficient implementation.
Alternative?
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Data description language example: XML
Some problems:
Unorthogonal design. Strange restrictions based on poor
understanding or inefficient implementation.
Ad hoc typing system
Alternative?
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Alternative?
Data description language example: XML
Some problems:
Unorthogonal design. Strange restrictions based on poor
understanding or inefficient implementation.
Ad hoc typing system
To know if an XML document is well-typed it is necessary to
prove some properties of an FSA which can be constructed
from the description.
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Alternative?
Data description language example: XML
Some problems:
Unorthogonal design. Strange restrictions based on poor
understanding or inefficient implementation.
Ad hoc typing system
To know if an XML document is well-typed it is necessary to
prove some properties of an FSA which can be constructed
from the description.
Notation for an empty list of elements?
Cookbook science vs. explanatory science
Examples
Reasons
What are the reasons for this situation?
Problems
Alternative?
Cookbook science vs. explanatory science
Examples
Reasons
What are the reasons for this situation?
Researchers’ genuine practical interests.
Problems
Alternative?
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Alternative?
What are the reasons for this situation?
Researchers’ genuine practical interests.
The funding system. The purpose of the research is to have
quick solutions to today’s problems.
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Alternative?
What are the reasons for this situation?
Researchers’ genuine practical interests.
The funding system. The purpose of the research is to have
quick solutions to today’s problems.
Confusion between development and research.
Cookbook science vs. explanatory science
Examples
Reasons
What are the problems with this situation?
Problems
Alternative?
Cookbook science vs. explanatory science
Examples
Reasons
Problems
What are the problems with this situation?
A lot of good researchers are wasting their time studying
things not worthy of studying.
Alternative?
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Alternative?
What are the problems with this situation?
A lot of good researchers are wasting their time studying
things not worthy of studying.
A lot of not-so-good researchers are wasting our money
studying things which are too complicated. There will be no
contribution to our common knowledge.
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Alternative?
What are the problems with this situation?
A lot of good researchers are wasting their time studying
things not worthy of studying.
A lot of not-so-good researchers are wasting our money
studying things which are too complicated. There will be no
contribution to our common knowledge.
Innovation and development is slowed down.
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Alternative?
What are the problems with this situation?
A lot of good researchers are wasting their time studying
things not worthy of studying.
A lot of not-so-good researchers are wasting our money
studying things which are too complicated. There will be no
contribution to our common knowledge.
Innovation and development is slowed down.
Good research has to be disguised. It is never good for ones
self-esteem to pretend to do things. Some examples:
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Alternative?
What are the problems with this situation?
A lot of good researchers are wasting their time studying
things not worthy of studying.
A lot of not-so-good researchers are wasting our money
studying things which are too complicated. There will be no
contribution to our common knowledge.
Innovation and development is slowed down.
Good research has to be disguised. It is never good for ones
self-esteem to pretend to do things. Some examples:
Denotational semantics is disguised as precise description of
Java or C.
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Alternative?
What are the problems with this situation?
A lot of good researchers are wasting their time studying
things not worthy of studying.
A lot of not-so-good researchers are wasting our money
studying things which are too complicated. There will be no
contribution to our common knowledge.
Innovation and development is slowed down.
Good research has to be disguised. It is never good for ones
self-esteem to pretend to do things. Some examples:
Denotational semantics is disguised as precise description of
Java or C.
Static analysis disguised as security.
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Alternative?
What are the problems with this situation?
A lot of good researchers are wasting their time studying
things not worthy of studying.
A lot of not-so-good researchers are wasting our money
studying things which are too complicated. There will be no
contribution to our common knowledge.
Innovation and development is slowed down.
Good research has to be disguised. It is never good for ones
self-esteem to pretend to do things. Some examples:
Denotational semantics is disguised as precise description of
Java or C.
Static analysis disguised as security.
Category theory disguised as UML.
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Alternative?
What are the problems with this situation?
A lot of good researchers are wasting their time studying
things not worthy of studying.
A lot of not-so-good researchers are wasting our money
studying things which are too complicated. There will be no
contribution to our common knowledge.
Innovation and development is slowed down.
Good research has to be disguised. It is never good for ones
self-esteem to pretend to do things. Some examples:
Denotational semantics is disguised as precise description of
Java or C.
Static analysis disguised as security.
Category theory disguised as UML.
We have to give up the strive for complete understanding.
Badly designed artefacts like Java, UML, XML leads to
scientific ”theories” which are too complex to be useful.
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Alternative?
What are the problems with this situation?
We should not forget the esthetics of science. A good theory
is beautiful. Looking at ugly things makes it difficult to create
a beautiful theory.
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Alternative?
What are the problems with this situation?
We should not forget the esthetics of science. A good theory
is beautiful. Looking at ugly things makes it difficult to create
a beautiful theory.
Good researchers want to work with beautyful theories.
Cookbook science vs. explanatory science
Examples
What should be done instead?
Reasons
Problems
Alternative?
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Alternative?
What should be done instead?
The minimum: When we study badly designed artefacts, we
should point out the problems and suggest alternative design
decisions. We should be more critical!
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Alternative?
What should be done instead?
The minimum: When we study badly designed artefacts, we
should point out the problems and suggest alternative design
decisions. We should be more critical!
Even better: Suggest new designs without trying to be
compatible with old ones.
Cookbook science vs. explanatory science
Examples
Computer Science education
Reasons
Problems
Alternative?
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Computer Science education
Cookbook science has a strong tradition in engineering
subjects.
Alternative?
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Alternative?
Computer Science education
Cookbook science has a strong tradition in engineering
subjects.
A funding system where we get paid for each point a student
makes is a strong impetus for cookbook teaching.
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Alternative?
Computer Science education
Cookbook science has a strong tradition in engineering
subjects.
A funding system where we get paid for each point a student
makes is a strong impetus for cookbook teaching.
The only way to educate engineers to a high standard is to
make them understand. This requires good theory. They
should use programming languages, description languages,
specification languages which can be understood. (I am not
saying that this exists yet).
Cookbook science vs. explanatory science
Examples
Reasons
Problems
Alternative?
References
References:
George Gale: Scientific Explanation,
http://cas.umkc.edu/philosophy/gale/CambridgeHistory.pdf
Peter Denning: Great Principles of Computing, CACM Nov
2003/Vol. 46, No. 11
Acknowledgements:
I want to thank John Allen for an email discussion on this topic.