HW4 HT2021

Homework 4 Panel discussions

Due Oct 11 17.00.

A panel discussion is a public debate in front of an audience. The panel consists of experts, chosen to reflect different opinions. The moderator keeps the discussion going by asking the panelists questions, moderating the discussion, and ensuring that each panel member has an equal opportunity to speak.

This is an example of a panel discussion:

You are invited to participate in two panel discussions in the next seminar. You should prepare for the following three roles:

• moderator
• panelist
• audience

You will not know beforehand which of two discussions you will participate in, so you must prepare for both.
Think of this exercise as role playing - it is allowed to argue passionately for standpoints that are not your own,
which may increase the entertainment value of the exercise.

You should prepare the panel debate through a written assignment:

1. The first panel discussion will deal with the subject ''Is Computer Science Science?"

1. Read the paper Is Computer Science Science?Links to an external site. [Peter Denning. 2005, Communications of the ACM Vol.48(4), pp. 27-31]. Other optional reading is listed below.
2. Write a short introduction (5-10 sentences) to this panel discussion aimed at a general audience.
3. Formulate two questions a moderator could ask the panel. Try to formulate them so that they provoke intense discussion.
4. Write down three arguments against calling computer science a science, and for each of them a strong counterargument.

2. The second panel discussion will deal with the subject ''Can AI algorithms be allowed to make sensitive decisions involving humans?''

AI systems increasingly govern people’s lives. In the US in particular, algorithms make decisions about health care, housing, insurance, education, employment, banking, and policing. Medical diagnosis systems based on deep learning have also made considerable progress, and in some cases outperform physicians. However, racial and gender biases, as well as other types of biases, can sometimes be deeply embedded in these systems.
This has generated considerable criticism and debate.  At the same time, people also suffer from bias and prejudices, so one could make an argument for the potential of AI systems to achieve a higher degree of fairness than humans.

1. Start by familiarizing yourself in more detail with one case of algorithmic bias, and summarize this case in a short paragraph
   with one or more references (some suggestions are given below).
2. Write a brief popular introduction (5-10 sentences) to the panel discussion.
3. Write down two questions a moderator could ask the panel. Try to formulate them so that they provoke heated discussion.
4. Choose a standpoint for or against, and write down three arguments for your standpoint, each with a possible counterargument.

Optional reading (you can of course also use any other material you find on these subjects) :

1. What is computer science? Selected opinions by some of the founders of the field.

Vinton G. Cerf, Where Is the Science in Computer Science? (Links to an external site.) Communications of the ACM, October 2012, Vol. 55 No. 10, Page 5.

Vinton Cerf (1943-) is an American Internet pioneer and one of the fathers of the Internet, together with TCP/IP co-developer Bob Kahn. They received the Turing Award (Links to an external site.) in 2004.

Newell, Allen; Perlis, Alan J.; & Simon, Herbert A. (1967), What is Computer Science? (Links to an external site.), Science 157(3795), 1373–1374.

Allen Newell and Herb Simon were pioneers in AI research during the 1950s, and received the Turing Award together in 1975. Herb Simon is also known for his contributions to other fields, includníng cognitive psychology and economics (for which he received the Nobel Prize in 1978). Alan Perlis did important work on early programming languages and compilers, and was awarded the first Turing Award in 1966.

Knuth, Donald (1974), “Computer Science and Its Relation to Mathematics (Links to an external site.)”, American Mathematical Monthly 81(4) (April): 323–343.

Donald Knuth is known, e.g.,  for his contributions to algorithm analysis and computational complexity, for his books The Art of Computer Programming, and for the creation of TeX. He received the Turing Award in 1974.

Brooks, Frederick P., Jr. (1996), “The Computer Scientist as Toolsmith II”,  (Links to an external site.)Communications of the ACM 39(3) (March): 61–68,

Fred Brooks is an American computer scientist who has done fundamental work in computer architecture, operating systems, software engineering, and human-computer interaction. He received the Turing Award in 1999.

And for those interested in reading further, there is an online draft of an excellent book by William Rapaport, a computer scientist and philosopher at SUNY Buffalo, called Philosophy of Computer Science  (Links to an external site.)(938 pages, but very readable). Rapaport is also known for pointing out that Buffalo buffalo Buffalo buffalo buffalo buffalo Buffalo buffalo  (Links to an external site.)is a grammatically correct sentence in English.

2. AI and algorithmic bias

A number of examples have dealt with racial injustice in the US. These include the work of Joy Boulamwini at MIT, 
see for example her TED talk How I am fighting bias in algorithms, Links to an external site. Links to an external site.The Algorithmic Justice League, Links to an external site.or if you have access to Netflix, the 2020 documentary Coded Bias. Links to an external site.

Links to an external site.And see also Buolamwini, J., Gebru, T. "Gender Shades: Intersectional Accuracy Disparities in Commercial Gender Classification." (Links to an external site.) Proceedings of Machine Learning Research 81:1–15, 2018 Conference on Fairness, Accountability, and Transparency.

Another researcher doing related work is Latanya Sweeney at Harvard, see for example Discrimination in Online Ad Delivery Links to an external site.

Additional examples of algorithmic bias include:

Amazon's abandonded recruitment tool Links to an external site. from 2015

Risk assessment tools in criminal justice Links to an external site.

On the other hand, some careful thought has also gone into how computer science and AI 
could be a force for fairness and social change, see for example the work of Jon Kleinberg
(a well-known theoretical computer scientist) and associates, e.g.,

Roles for Computing in Social Change Links to an external site.
Algorithmic Fairness  Links to an external site.
and other publications.

Some examples of how machine learning is beginning to out perform doctors in 
limited medical diagnosis tasks:

How AI is improving cancer diagnostics Links to an external site.
Artificial intelligence is improving the detection of lung cancer Links to an external site.

Risks and benefits of an AI revolution in medicine Links to an external site.

(more examples may be added later, including arguments for AI)

Handing in your solution

Please save your solution as a pdf file and hand in in Canvas (for grading). There will be no peer
review of Homework 4!

Feedback from your TA

Your seminar leader will grade your submission and report the result in Canvas. This may happen before
the associated seminar, but if your seminar leader is busy it will 

Complete means you have passed the assignment.

Incomplete means you have to hand in a revised version.

Fail means that you will have to submit a new version and attend the make-up seminar.
The Fail grade will only be applied in exceptional circumstances such as plagiarized work.