The Honor Code and discussion of homework assignments

[Morgaine]

There has been a fair bit of admonishment and even accusation on Reddit and in Google Groups about the depth of discussion of homework assignments.  It's clear that views differ somewhat on how the Honor Code is to be applied to Internet discussions, so I looked for authoritative advice at https://www.ai-class.com/overview in the section 'Discussions' --- I've highlighted some key phrases in bold:

Discussions
There will also be a general discussion forum, in which you can discuss questions and interact with other students. You are not allowed to post solutions to active homework assignment and exams here, but you are allowed to discuss the material covered in class; and you can of course pose questions. Once the answers to a homework assignment have been posted you are free to discuss them, as well as sharing any code you may have written.

So, the boundaries are very clear and you can't go outside them, but just inside the boundaries there is still a surprisingly large undefined region.  Posting solutions is completely disallowed, but there is no express mention of discussing solutions, nor of discussing the problems themselves.  This is not at all surprising, since the professors very strongly encourage collaborative study groups as well as various forums for open Internet discussion, and there is no possible way of making some discussions OK and others not.  It's a continuum.

Personally, I think the most important word of all in this area is Honor.  It is on your honor that when you do an assignment (and of course the exams), you will put aside all your connections, all your friends and acquaintances and friendly neighborhood AI experts, and you will work ALONE on the material, using your own memory and your own understanding and nothing else.  Because that's what assessment is all about, quantifying your own, personal competence and understanding of the topic.

But until you choose to start doing assignments, I expect that the professors are deliberately placing no limits on the extent and depth and coverage of discussions, and even want to encourage them to the greatest extent possible because that strengthens learning.  There is no learning occurring when assignments confuse students and the questions are answered semi-randomly.  I am certain that the assessment is intended to quantify the degree of learning and to enhance it, and not to assess the quality of the videos.  This is probably why the homeworks videos are posted so far ahead of the deadline, to provide time for effective discussion and the learning that comes with it.

Unless there is some additional official guidance forthcoming, I suggest to focus on the word "Honor" and take it to heart honestly by working alone when you open the assignments screen.  It's what you personally learned that matters, not how you did the learning.


Morgaine.



===================

On Fri, Oct 14, 2011 at 10:42 PM, Graeme Humphrey <graeme...@gmail.com> wrote:

@Bruce I hadn't considered it to be and I hope you are not accusing me
of cheating. I thought the question was designed for me to prove my
understanding of BFS and DFS algorithms. If I was cutting code to do
this I would have some logic to pick evenly ranked frontier nodes, and
the outcome, because they are evenly matched would be a moot point.
Each of the nodes are all candidates to be  expanded and we could
quite as easily pick them from a hat.

I was NOT asking for any help in deriving the answer to the question,
I was only asking for help in defining the question and thought it was
useful for discussion.

On Oct 14, 2:21 pm, Bruce Seely <seel...@gmail.com> wrote:
> Isn't that part of what you are supposed to be figuring out as part of
> the homework?
>
> On 10/14/11 5:19 PM, Graeme Humphrey wrote:
>
>
>
>
>
>
>
> > I'm doubting my logic in Q6 now. We are told our choice of equal
> > frontier nodes is to be made left to right, or vice versa. That's all
> > fine and dandy when traversing down a tree in one direction. But what
> > happens when you start going back up in the opposite direction. Are
> > our node choices still the same? If it was solely a decision of me
> > looking at the graph to ascertain which is left or right, it would be
> > an impossibility as some child nodes are aligned vertically. So my
> > assumption was that the "left" and "right" changes as my heading
> > changes.
>
> > Any thoughts?
>
> > On Oct 14, 12:35 pm, Graeme Humphrey<graeme.hu...@gmail.com>  wrote:
> >> After re-watching the video and reading the text, I'm reasonably
> >> confident that "expansion" occurs when the node is chosen, removed
> >> from the frontier, and the child nodes are placed on the frontier.
>
> >> On Oct 13, 2:29 pm, Graeme Humphrey<graeme.hu...@gmail.com>  wrote:
>
> >>> My thought (and hope as this was how I did it) is that their
> >>> definition of "expanded" is visited. I'm mildly confident of this
> >>> because the last question (A* search) refers to sequential nodes being
> >>> expanded, rather than a set of nodes being expanded. However the last
> >>> question differs as they do not include the root node in the count, so
> >>> who knows.
> >>> On Oct 13, 2:05 pm, anthonyh3<anthon...@gmail.com>  wrote:
> >>>> With regards to Homework 1, has anyone gotten any clarification as to
> >>>> what "expanded" means when talking about nodes in Search Tree and
> >>>> Search Tree 2? I was hoping someone would have "official"
> >>>> clarification...
> >>>> Does "expanded" mean expanding the nodes attached to the current
> >>>> state, or does it mean visiting the current node (state)? There could
> >>>> be a big difference in answers based on the definition of expanded.
> >>>> Consider:
> >>>>       A
> >>>>   /    |   \
> >>>> B   C   D
> >>>> Assuming B is the goal state, and the search is left-to-right, if
> >>>> "expanded" means expanded, when node A is the current state, nodes B,
> >>>> C, and D are _expanded_ and placed on the frontier. Next, node B is
> >>>> visited, reaching the goal state and stopping the search. 4 nodes have
> >>>> been expanded when the search stops. In contrast, if "expanded" means
> >>>> visited, then only 2 nodes would be expanded when the search stops, as
> >>>> only A and B were visited.
> >>>> Based on the simplicity of the trees in question for these homework
> >>>> questions, I am going to assume that "expanded" actually means
> >>>> visited, otherwise, 3 of the 4 searches would have the same answer.
> >>>> Thanks for any clarification anyone can give.

--
You received this message because you are subscribed to the Google Groups "Stanford AI Class" group.
To post to this group, send email to stanford...@googlegroups.com.
To unsubscribe from this group, send email to stanford-ai-cl...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/stanford-ai-class?hl=en.

[David Weiseth]

Those that take the shortcut on the homework, they may find the exams crushing.  You can review the homework following its submission and grading, but the taking of the prep exams a.k.a. homework is to prepare you for the crush of the exam.

The bottom line is to become educated, but the "advanced track" offers the additional credential that you were able to run with Stanford level student work.  

This in the end is an experiment of the grandest scale, and I laud both Peter and Sebastian for embarking on this, this is truly wonderful stuff and the whole field of A.I. will benefit.

Cheating will always occur even with normal classwork, but it does make it easier to find folks eager to post their prowess even if in the end this reduces the viability of the "advanced track" credential.

I think if I was Peter and Sebastian, I would clamp down on the exams, rather than the homework.  I would give people windows for taking the exam to prevent overloads on the servers.  This could be an assignment one week prior by random or first come first serve.  People can petition for a change within 48h for hardship, but otherwise it stands.

Again the growing pains are well worth the result.  --David