Many of you are rapidly approaching the end of your degrees. Unfortunately, too many of you have not yet had the opportunity to
a) read good, scholarly literature, and
b) write good, scholarly literature.
Both of these skills are extremely beneficial as a computer scientist. Many of you may still believe that you will get jobs where all you do is write code. The fact is, even programmers spend a vast amount of time involved with writing and communicating ideas. Programmers write problems descriptions, program designs, program and end user documentation, grant proposals, reports, papers, articles etc. The ability to talk and write about your knowledge is probably just as important as actually having the knowledge itself. I was fortunate, that my advisers in graduate school required that I hone my reading and writing skills before I was "allowed" to do "real research." I now know that these skills have absolutely opened doors for some incredible opportunities.
Therefore, in addition to any writing you may do on homework assignments, you will write one large research paper this semester. This paper will be written as though you were submitting it for publication at the annual conference of one of the top AI organizations - the American Association for Artificial Intelligence (www.aaai.org). Although the topic of your paper will be different from what is normally submitted for AAAI, the process you follow and the requirements you work within, will be very similar to those of an AI researcher preparing for publication at AAAI.
The goals for assigning a big paper include:
I hope that you will view this paper not as a burden but as an opportunity to explore a topic that interests you and to practice your professional writing skills.
As I alluded to above, the papers that are normally submitted to, and published at, the annual AAAI conference are formal research papers (you will have the opportunity to read one or more of these papers during one of the homework assignments this semester). These are papers which discuss the results of a specific research project conducted by the authors and will normally involve the construction of a system, the testing of the system, and a discussion of the results of this system compared to some set of benchmarks.
Clearly, you are not going to know enough AI to conduct this type of research. Instead, you are going to perform literature based research. You will analyze approximately a dozen different books, magazine articles, and websites to write a paper on an AI topic that interests you. The actual format for the paper you write will likely fit into one of three types of paper:
A survey defines a topic, identifies the interesting questions in the area, and then describes the different kind of work that has been done to seek answers to these questions. For example, you might select the topic of AI and the law and write a survey of legal reasoning systems that use AI.
The emphasis of a survey paper is on the description of what has been done. Your contribution as survey author lies in how you organize the questions and how you describe the work.
A tutorial aims to teach the topic to readers with sufficient general background who lacks specific background in the topic. Such a paper presents the important concepts in the area, relates them to one another, and identifies open questions for further study. For example, you might select the topic of case-based reasoning and write a paper that would help your classmates learn the topic without having to read the whole literature.
The emphasis of a tutorial paper is on teaching the concepts of the topic, almost like a textbook chapter. Your contribution as tutorial author lies in what concepts you select, how you organize them, and how you present them to the reader.
A problem analysis aims to describe a specific problem and to describe how it might be solved (in this course, with AI techniques). Such a paper defines the problem in the vocabulary of the problem domain, considers one or more AI approaches that could be applied to the problem, and then selects one approach and explains why it is the best or most promising. For example, you might select the topic of AI in music and do an analysis of how a computer program might compose a symphony.
The emphasis of a problem analysis paper is on the problem being studied. You discuss and analyze AI techniques and tools in the context of just that problem. Your contribution as analyst lies in your presentation of the domain, your careful presentation of options and trade-offs among them, and the suitability of your recommended technique.
You may propose to do a different sort of paper, but if you do you will need to justify both the topic and your paper format.
Any topic within AI is fair game, as is an application of AI to any problem domain. Select a topic from your own experience and interests, and you will find the project more engaging. The result will be a more interesting -- and, usually, better -- paper.
If you decide to look for a topic in the text or lectures, identify a problem or issue that goes beyond what we read about or discuss. For example, if you find that you really like the area of planning systems, you might select an open question in planning, such as how best to re-plan in the face of a changing environment, and write a survey of techniques for solving the problem. Or, if you find that you really like the topic of learning, you might select a learning technique that we don't discuss in great detail and write a tutorial.
Here are some topics that interest me. If one of them interests you, consider writing your paper on it.
Due : Friday, September 7th at the start of class.
Background : While you won't conduct bleeding edge research in this paper, you will consider a more weighty problem than you are likely used to tackling. There will be two main challenges in writing this paper. One, reading previously published research, some of which may start out over your head. Two, learning how to write a short and concise paper when you, likely, have a lot to say. This assignment aims to help you begin to tackle these challenges by "forcing" you to read and analyze what is probably your first AI journal article. In doing so, I hope that you will begin to understand some of the challenges earlier rather than later.
Deliverable Specifics : For this assignment you should select a research paper about AI from one of the following sources:
If you have any doubts about the acceptability of a paper you find, please discuss it with me.
While it is not required that this paper have anything to do with the topic you are beginning to consider for your research paper, it would certainly make sense, since this paper could be one of your first sources for this paper.
Read the paper, and then write a short response to this paper (2-4 pages, 600-1200 words). Your response should explicitly address the following questions.
You should submit a hard copy of your response at the start of class on the due date. You should also submit an electronic copy to me via the homework submission system.
Due : Friday, September 28th, at the start of class
Background : Often times, before you submit a paper for publication, the editors or organizers will issue a CFP or "Call for proposals." While the exact format of these proposals will vary, they typically involve having you write a relatively brief but concise description of what it is that you would like to write about in your full paper. By having this short proposal approved by the editors in advance, you know that you won't spend weeks working on a paper that is deemed completely irrelevant for the conference. While acceptance of the abstract does not guarantee acceptance of the final paper, rejection of the abstract tells you that your paper is targeted towards the wrong conference or journal.
Similarly, for this class, you will write a short proposal so that you don't waste several weeks researching a topic that is deemed inappropriate for this course.
Deliverable Specifics : By the start of class on the due date, submit a hard copy of your proposal. You should also submit an electronic copy to me via the homework submission system.
This should be a 400-500-word description of the topic you'd like to explore and the type of paper you'd like to write. I will try to help you make sure that you have a topic of the right scope and that there is sufficient material available for your research. (I will also try to ensure that I don't end up with 20 papers on the same topic!)
Notice that by this point in the semester, we will have covered less than half of the actual content of this course. In order to select a topic, you may want to skim ahead in the text to learn a bit more about what topics are available. Feel free to discuss any ideas you may have ahead of time with me. I'll try to point you in a good direction.
Due : At the start of class on Friday, November 9th.
Background : The next step in submitting a paper for a conference or a journal is the submission of the first complete version of your paper. This paper should be one that you have written and rewritten several times. You should feel that it shows the quality of work that you are capable of producing. At a conference like AAAI, the organizers will typically receive 2-5 times as many papers as they actually have the ability to accept for publication/presentation. At this stage, authors submit what they consider to be their best work - not some kind of rough draft - and a peer evaluation is conducted. Each paper would be read by 2-4 peers - other researchers who understand the type of work being conducted - who evaluate the paper on the quality of research being conducted, the impact of the results being discussed, and the readability of the overall paper. In the end, these peers will advise the editors/organizers as to whether or not the paper should be accepted for publication and only the top vote getters will be accepted. Thus, it is crucial that the paper that authors submit is complete and well written.
While each and every one of you will have your papers "accepted" we still want to have you put some serious effort into this first full version of your paper. Notice that I do not use the language "rough" draft or "working" draft. I have used the language "review" version. In other words, this paper should largely be a complete paper that you would consider submitting for a final grade in a class. Your peers will be editing this paper and trying to help you improve it before you submit the final version, but they will also be evaluating your paper, and you don't want them telling you (and me) that what you have so far isn't very good.
Deliverable Specifics : You will submit three (3), separately stapled copies of your paper. One of these will go to me. The other two will be randomly distributed to two of your classmates who will perform a peer evaluation on this version.
You should also submit an electronic copy to me via the homework submission system.
The paper should be written in the style of a AAAI conference paper using the following rtf file for AAAI publications as a starting point.
FORMAT:
| "Figure 1. The module structure of the fish-finding expert system.". |
Table captions lie above tables; figure captions lie below figures. The text should refer to each table or figure by its number: "Figure 1 illustrates the system's structure." Table and figures may be two columns wide, if necessary; single-column tables and figures must fit within their columns.
Unless we discuss otherwise in advance, your draft must follow from the topic we agree on after you submit your proposal.
Due : At the start of class on Friday, November 16th.
Background : An important part of being a member of a research community is the participation in peer reviews. In this activity you are given one or more papers authored by other people, and are asked to read and evaluate the papers, often following a very specific format for evaluation. Additionally, you are typically asked to provide very specific comments on not only simple things like grammatical errors ("you have the word it's when you meant the word its") but more advanced things such as issues of confusion ("I don't understand what you mean by the term 'nearest neighbor' in the paragraph about alternate methods") or suggestions for improving the paper ("The paper might flow a little better if you moved the section on motivation before the section on prior work. Of course, doing so will require rewriting so that you don't refer to concepts introduced in the prior work section"). Being a thorough and helpful peer reviewer is an important skill as well.
Deliverable Specifics : For each of the papers that you were asked to review, print and complete the following form. Please provide specific and helpful comments for each paper. Note that I will be assigning a grade to your evaluations based on the amount of effort that it appears that you put into providing helpful comments for your classmates.
Due : At the start of class on Monday, December 3rd
Wednesday, December 5th.
Background : After peer reviews are performed, the editors/organizers will decide which papers to accept and which ones to reject. Authors are notified which category they fall in, and the peer reviews completed on each paper are returned to the original authors. If a paper is rejected, peer reviews should help the author understand why. If papers were accepted, they give the author some ideas regarding what the reviewers liked and didn't like about the paper. Authors use these reviews to improve the quality of their final paper. In fact, for journal articles there often is a period of time where the authors must provide the editor with a document indicating how they specifically addressed each of the reviewers concerns. Most of the time this involves making the change the reviewer suggested, but it can also include defending how the authors chose to do it in the first place ("While we understand the reviewer's comment on including a section about "big challenges" we feel doing so changes the tone of this paper.").
Deliverable Specifics : You should read through the comments provided by each of your reviewers. In most cases, you will be making the changes suggested in your reviews. In certain cases you may choose to ignore comments, but you should not do so without making it a conscious decision. Put the time into making sure that your final paper reflects a semester's worth of research and writing.
At the start of class you should submit a paper copy of you research paper. You should also have used the homework submission system to submit:
Grades will be determined on the basis of content and form.
I really do care about the quality of your writing. Please use this as an opportunity to practice and to learn.