HOW TO PROGRAMME CHAIR IJCAI Alan Bundy 29.7.83 These notes describe how to be the programme chairperson of IJCAI, or at least how I think it should be done based on my experience of doing it for IJCAI-83. They should be read in conjunction with the various files, electronic and paper, that I accumulated during that period, and which provide more detail. Many of the procedures I adopted are only relevant for a conference the size of IJCAI, i.e. about 500+ submitted papers. A discussion of the particular problems of handling such a large conference can be found in DISCU1-3.MSS, PLANS.TXT and ARTICL.TXT. The notes are organised around the various task that the programme chairperson has to do, with an indication of when they must be done by and how long they will take, and which files give what relevant detail. KEY Programme - list of events, performers, etc. Program - computer software Programme Committee - committee of which Programme Chairman is chairman. IJCAI - International Joint Conference on Artificial Intelligence. Name of each conference. IJCAII - International Joint Conferences on Artificial Intelligence Inc. Name of company which runs IJCAIs. Organising Committee - committee which runs IJCAI, consists of General, Programme and Local Arrangments Chairmen and Secretary/Treasurer. Trustees - committee which runs IJCAII. consists of General and Programme Chairmen, Secretary/Treasurer and three other members, usually the last three general chairmen. Conference Committee - committee of which General Chairman is chairman. Executive Committee - committee which meets at nth IJCAI to decide officers/venue for n+1st IJCAI. File conventions: 'paper', ELECTRONIC. GENERAL POLICY When I took on the job I was given a list of identified problems with the way IJCAIs had been run in the past and asked to think of ways of solving these problems. The problems and my attempted solutions can be found in DISCU1-3.MSS, PLANS.TXT and ARTICL.TXT and in 'General Policy'. The paper file contains my discussion documents, feedback to them, unsolicited letters complaining about previous IJCAIs and advice from previous programme chairpeople: Bruce Buchanan, Ann Drinan (Schank's assistant), and David Waltz. I described my final plans in an article circulated to several AI newsletters. It is ARTICL.TXT. It took from September 1981 to May 1982 to come up with a general policy agreeable to the trustees and other relevant people, and I did not solve all the problems. Things I will admit are still wrong are listed in MISTAK.TXT. That was not the end of it, since the necessity of discussing every major policy decision with the other members of the conference committee at least doubled, if not trebled, the amount of work required to programme chair and the length of time required to do it. For this reason, I strongly urge my successors to use my solutions as a basis. In practice, the people who gave me most feedback, and with whom I established the strongest working relationship for day to day decision making were the organising committee. ADMINISTRATION You will need: an administrative assistant, a photo-copier, a typewriter, a phone for international calls, copious stationery and cabinets, and access to the ArpaNet. The assistant will need to work about 9 person months, but the workload is very uneven, with very little during the first year, and with big bursts of activity around the deadlines. So find someone who is willing to work such hours, and who is reliable, conscientious, keen, able to type and use a computer, and is a good administrator. Recruit someone early so they can have a say in early decisions, even though they may not be employed then. I found that a finishing postgrad who needed the money, was ideal. The quality of the person you employ is the major factor determining how much of a nightmare this job is going to be. I was very lucky. To help you both cope with the big rushes, you will want money to employ casual labour: students and secretaries. Make a budget and revise it constantly. Previous examples can be found in file 'Budget Statements'. Since you are only going to do this once (believe me), it is probably not worth investing in computer programs and databases to store your information. CO-SPONSORSHIP Though not strictly part of my job, I agreed to help find co-sponsors for the conference - both academic societies (who offer help in kind) and industrial labs (who offer money). Details are in 'Co-sponsorship'. RECRUITING PROGRAMME COMMITTEE On my scheme you want one member, called the subfield chairperson, for each subfield of AI. You want people of some weight in these subfields, people who have a sense of responsibility, and who will be easy to contact, e.g. via netmail, but with some international spread. These criteria are antagonistic, to some extent, and you have to compromise. Since I was not qualified to judge the merit of candidates in many subfields, I asked for suggestions from other members of the organising committee and the trustees. It took much longer to recruit the programme committee than I had anticipated - 5 months! About 50% of the people asked declined, usually through pressure of work. Given that the further ahead people are asked to make a committment the more likely they are to say yes, I think you should begin recruiting immediately. The duties I expected of subfield chairpeople are listed in DUTIES.TXT, which I used in my recruiting drive. The committee I recruited and the subfields I used are in COMM.MAI. Sample recruiting letters are in 'Recruiting Programme Committee'. My initial plan called for subfield chairpeople and referees to be clumped in geographic centres of excellence. A personal list of such centres is in PLACES.TXT. However, in the age of electronic mail and telephone, I no longer think this criterion is necessary. Nor is it desirable, since it leads to one sided refereeing. Nor is it possible, since there are too few big centres. The one major item of policy you must have decided before recruiting is how to divide AI into subfields. Some discussion of this is in the paper file. The only papers I had difficulty placing were a survey of AI in China and an education one, but neither justified a subfield of its own. RECRUITING REFEREES Under my scheme referees are recruited by each subfield chairperson. Only high quality, reliable people should be recruited, representing all the major methodological positions. Referees are usually prepared to take about 5 papers each. Estimate how many you will need from the statistics given in: NUMBER.TXT, STATS.TXT, STATS.RNO and 'Statistics'. Make plenty of allowance for an increase of submitted papers and a different distribution of papers to subfields. I got caught on this and ended up overloading referees. The recruiting process takes many months, so start the subfield chairpeople doing it as they themselves are recruited. The duties required of referees are listed in REFERE.LPT. The circular I used to help subfield chairpeople work out who to recruit is in GETREF.MAI. I needed to follow up with REF.MAI to clear up misunderstandings. The Yale blacklist mentioned there is a list made during IJCAI-81 of referees whose reports did not arrive in time (possibly due to postal delays). I checked all suggested referees against it. It is in BLACK.TXT and 'Referee Recruiting' and should be treated as highly confidential. In fact, we used some people on the black list, and they were fine, so anyone who is on the referee list should be deleted from the black list. After the referees have been recruited unofficially, they will appreciate an official, circular letter from you confirming their recruitment, e.g. LETTER.TXT. This also incorporates a reply card for supplying keywords, which are invaluable when it comes to allocating papers to referees. These documents are duplicated in 'Referee Recruiting'. BROCHURES You will need to produce 3 brochures for general distributution: the call for papers, the 1st brochure announcing the conference, the 2nd brochure describing the programme and events. The actual printing and mailing is handled by the local arrangements chairperson, but you will need to provide the text. The call for papers should go out as soon as possible, and not later than six months in advance of the deadline for submissions. Allow plenty of time - from first draft to distribution took me 5 months! It should be submitted to every AI journal and newsletter in sight, and this can be done before the printed version is ready. Expect some journals to take months, if not years, to publish it. The call will incorporate many of your general policy decisions, e.g. the division of the field into subfields, the timetable/deadlines, the types of paper acceptable, etc. My call for papers is in CALL.TXT. If you have recruited them in time, it would be good to include the programme committee - I had not, so did not. The 1st brochure is the concern of the local arrangments people. It is mostly an advert for the conference. They will probably want some words on the programme. Mine are in BLURB.MSS. The 2nd brochure goes out to registered participants, authors, etc, just before the conference and/or at the conference. Its main purpose is to describe the schedule, but it also contains announcements of prizes, welcoming addresses, maps, local information, etc. Produce it as soon as you have the schedule fixed, and you will avoid having to field lots of questions of the "when am I on?" variety. My entries for this were: PROGRA.MSS (programme notes), SPECIA.TXT (about special sessions), SPRING.TXT & CANDT.TXT (prize announcements). In addition to this brochures it spreads awareness of the conference if you put major announcements in AI newsletters and on the ArpaNet-BBoards@MIT-MC. Suitable cases are: an initial conference announcement, the call for papers, prizes and awards, deadlines, changes and errors. Examples of publicity material can be found in 'Publicity'. INVITED TALKS AND PANELS I had two types of invited talks and two types of panels - general and specialized. The specialized talks were called keynote addresses. The arguments for this policy are given in PANEL.TXT and KEYNOT.MAI and the final system is described in TALKS.TXT. This policy could do with reviewing in the light of experience at IJCAI-83. I asked for suggestions from the everyone I could think of for invited talks and general panels, and picked the best 4 talks and 3 panels having regard for balance across subfields, burning issues, who had talked in the past, etc. The keynote addresses and specialized panels were subcontracted to the subfield chairpeople. The details for IJCAI-83 are given in KEYNOT.TXT and the panelists for IJCAI-79 are given in PANEL.LET. This procedure can be implemented as soon as the committees are all formed. Plenty of time should be allowed for panelists to be recruited by the panel organisers. IJCAI-83 subsidised invited and keynote speakers, panel organisers and subfield chairmen by advancing up to $500 towards travel and subsistence ($400 for people already on the right continent) and giving other panelists free registration. Industrial co-sponsors also got free registration. All conference expenses of the programme chairman his assistant and other members of the organising committee, were also met from IJCAI-83 funds. Most of this material and standard canvassing and invitation letters can be found in 'Invited Talks etc'. PRIZES AND AWARDS IJCAI-83 had four prizes/awards: the Computers and Thought Award, the Distinguished Service Award, the Springer-Verlag Prize and the Program Verification Prize. Only the Computers and Thought Award is guaranteed for subsequent IJCAIs. The Distinguished Service Award is only given if a suitable candidate emerges. Springer-Verlag may decide to give further prizes. Woody Bledsoe's committee may come up with another program verification prize, or a prize in another area. There have been prizes for chess and theorem proving in the past. A lot of discussion about these prizes can be found in 'Prizes'. The Computers and Thought Award is your business. The procedure is described in CT.TXT and criticised in CANDT2.TXT. As a result of the latter I hope that the vaguenesses and ambiguities will soon be cleared up, and a tighter document will soon be issued by the Trustees. You must call for nominations, distribute these and the ballot paper (see CT.TXT and CTFORM.TXT), add up the votes, announce the result, and get a certificate made and presented. The announcement should be made in the ArpaNet-BBoards, AI newsletters, and the 2nd brochure (see CANDT.TXT). The Springer-Verlag prize, if it exists, should be handled in a similar way, except that the franchise is restricted to the programme committee. The original announcement of the prize can be found in SPRING.MAI. The mechanism for judging it can be found in SPGMEC.TXT. The announcement of the winner can be found in SPRING.TXT. Lots of discussion about the mechanism, criteria, principle, etc can be found in 'Prizes'. The Distinguished Service Award is handled by the General Chairman and the Trustees, and the program verification prize by Woody Bledsoe. Your only responsibility is to get a certificate designed and printed. The wording of all four certificates can be found in CERT.TXT. REFEREEING PROCEDURE I asked for four copies of the submitted papers. One copy went to each of the two referees, one to the subfield chairman, and one was kept for my records. Referees were selected by matching their keywords against the abstract of the paper, and by trying to pair them so they could discuss the paper. They did this face to face or by phone or net. After the two referees had discussed the paper they sent their reports to the subfield chairman, who made the accept/reject decisions and sent these and the reports to me. I confirmed the decision and sent an accept/reject letter, together with the author's section of the referee's reports, to the authors. Each person in the chain was supposed to photo-copy the reports before entrusting them to the mail - since some were bound to go astray (and did). The whole process needed to be completed in a couple of months, and this required lots of pushing and pulling from me. This procedure is given in more detail in REFERE.LPT and discussed in the policy discussion documents and REFPRC.MAI. Note that it is a continuous stream process and did not involve a programme committee meeting. The referee form I used is in REF.TXT, and various accept/reject letters in ACCPTL.TXT, ACCPTS.TXT, REJECT.TXT, REJECT.LET, ACCPTC.LET, ACCPTL.LET. I also sent authors a reply card, REPLY.TXT and some instructions, AUTHOR.TXT. You can find the instructions sent in 1979 in AUTHOR.LET and SLIDES.RNO. 'Referee Forms' contains examples of all the standard forms and letters from IJCAI-83, plus some examples from previous conferences. 'Refereeing Procedure' contains discussion of: deadlines, mailing, word counts, criteria of acceptance for long and short papers, referees conferring with authors, promotion/demotion of papers, movement to other subfields, multiple submissions, choice of subfield, anonymous refereeing, emergency procedures, proxy presentation of papers and other aspects of the refereeing procedure. After the last accept/reject letter has been posted off, and you breath a big sigh of relief, the 'your referees are blind' complaints letters will start coming in, so be prepared. Typical complaints are (apparent) lack of consistency between referee forms and final decision, not enough detail given of grounds for refusal, etc. See my 'Complaints' file for further detail. Some problems might be avoided by further instructions to referees on standards required and how to fill in forms. FIXING THE SCHEDULE Once you know which papers have been accepted you can start making the schedule. In my scheme short and long papers were not mixed in the same session. I asked the subfield chairpeople to fix the schedule in their area, and then (with the aid of a student and a previously worked out outline, TIMING.TXT) put these together, dovetailing long sessions, but not short. The results are in SHORT, LONG, SESSIO.TXT and 'Schedule'. We sent a reply card with each acceptance letter, so that authors could let us know whether their paper would be presented, by whom and what special audio visual equipment was needed. This information was useful in keeping the schedule up-to-date and advising session chairpeople of the details of their session. The audio visual needs should be communicated to the local arrangements people as soon as possible (we left it too late). PROCEEDINGS Most of the work involved in producing the proceedings has now been taken over by Kaufmann, the standard publisher of all IJCAI and AAAI proceedings. This deal is described in 'Proceedings'. They will: send out model paper and instructions, receive model paper from authors, work out indexes, contents pages, etc, put in page numbers, and get the proceedings printed. You must supply: names and addresses of accepted authors, structure of proceedings, foreword, lists of committee members, referees, etc, and other preliminary material, and front cover design. The structure of the IJCAI-83 is in PROCS.TXT. I was asked to make this a model for future proceedings. You are not bound to this structure, but if you improve it then please supply a revised structure for your successor. One area that requires further thought is the order of papers. The IJCAI-82 papers are grouped by subfield and then by session. I had mixed subfield long paper sessions and this tended to mess up the contents page. There may be no better way to do it though. There is a discussion of alternatives in 'Proceedings' and PROCS .MAI. My foreword is in FOREWO.TXT and the order of papers sent to Kaufmann is in the files LONG AND SHORT. 'Proceedings' also contains discussions of: fount size, author names, page charges and cover design. RECRUITING SESSION CHAIRPEOPLE When the schedule has been fixed you need to recruit session chairpeople. As usual, I left this too late, and had to run about at the last minute. Expect that 50% of the people you ask will not be able to do it because they are not coming to the conference or not on the right day. 50% of these will never reply to your invite, so put a deadline/cutoff in it, or you may end up with no chairperson or more than one for some sessions. Asking only authors of accepted papers ensures a higher hit rate, but even this is not guaranteed. You want people with an established reputation in their subfield who are able to chair a meeting, keep people to time, etc. You also want some balance by nations, sexes, etc. I asked the programme committee for suggestions in subfields not familiar to me, but they will not know who is going to the conference. The instructions I sent to session chairs are in CHAIR.TXT, and the ones from 1981 are in CHAIR.INS. FRINGE ACTIVITIES There are various fringe activities at IJCAIs, outside the main body of contributed and invited papers and panels, e.g. program demonstrations, industrial exhibition, book displays, special interest sessions, free sessions, posters, films and videos. At IJCAI-83, the first three of these were handled by the local arrangements people. The details of the special interest sessions are given in SPECSE.TXT, SPECIAL.TXT and 'Fringe Activities'. These are a good vehicle for dealing with people who write in wanting to do something with their chums in "fuzzy logic for pattern recognition". Free sessions were handled by making a room available for several slots, putting up a booking form, and letting the last person signing the form be the chairperson. Posters and films/videos I did nothing about and probably should have. Just before IJCAI-83 a desire emerged, both among concerned members of the public and the AI community, for an open (to the public) discussion on the social implications of AI. What details I have are in 'Public Panel Discussion'. Although the intellectual content of this discussion was not high, this probably reflects the widespread ignorance about the issues. I would welcome such open discussions becoming a regular part of IJCAIs, but I recommend trying to focus them on one aspect of the problem. Good Luck - hope you don't regret saying yes, too much.