Inf1B Proposal - comments by P.Stevens
=====

- I am not reassured by the further detail, especially as regards the
  integrity of the assessment. For example, because:

-- When I took over Inf1OP, it used to have lab exercises that were
   submitted for a small amount of credit, as proposed here. At the
   end of the course I looked at the marks, and also at the code,
   comparing coursework with exam. It was very clear that *many*
   students had, somehow, submitted coursework exercises containing
   correct use of language constructs they had *no idea* about how to
   use in exam. In other words, such submission does not effectively
   check student learning. It does teach students they can cheat with
   impunity.

-- MOSS similarity detection software finding similarities, among 400
   (or even 100) submitted solutions to a short exercise suitable for
   beginners, will not reliably detect cheating. Even trivial changes
   will blur the line between "these students copied" and "these
   students took similar approaches to an easy question" to the point
   where we can prove nothing.

-- Indeed, in this course it is hard to see how you would even explain
   to well-intentioned students exactly what constituted plagiarism.
   How will a student who struggled to get their code to work, and
   eventually succeeded with help from teammates or others, even know
   whether the degree of help they have had is allowed? If they think
   on reflection that they have had too much help, because they
   couldn't get their code to work without it, then what should they
   do?

-- Moreover, MOSS is no defence whatever against purchased solutions.
   See for example
   https://www.fiverr.com/gigs/java-assignment
   (Prices start at £4.08)

-- "Require students to explain their code and answer questions during
  the demonstration" -- i.e. in the 20 minute "group demonstration of
  the running code to a tutor". But most tutors will themselves be
  students - maybe even undergraduate students, certainly including
  MSc students who are with us for only one year. It is simply not a
  reasonable expectation that they will be either able or willing to
  identify students who have not written their own code in such a
  setting.

-- Note also that the online code-selling services often include
   detailed code comments and/or separate explanations, so that a
   student seems to understand how their code works is no guarantee
   that they wrote it.

-- As many investigations have found, it is *not* only a small
   minority of students who will cheat given the chance. Even if you
   don't care about it in this course (and you should, because we will
   all suffer if students pass a programming course without learning
   to program) you should care about the general lesson we're teaching
   students, i.e. that we don't much mind them cheating.

- The proposal is to spend more student time on learning OO
  programming, and to have correspondingly expanded learning
  objectives. Proposers argue that these expanded LOs require
  assessment methods beyond exam. Fine. But what is the argument that
  programming exam should not be used *in addition to* whatever else
  is done, to ensure that all students do learn basic programming? I
  would be considerably reassured if every student who passed this
  course had definitely demonstrated that they can program
  independently under exam conditions, to at least the standard
  required to pass Inf1OP. That should be a low bar.

- Moreover the resourcing required here is a major concern. Even if it
  is possible to recruit the extra TAs/tutors/markers required, that
  will surely be at the expense of other teaching support roles. (I
  say this as someone currently struggling to teach a 20 point course
  alone, with two of the three agreed teaching support positions
  unfilled for lack of applicants.)