There is a happy ending to a time-consuming quest reported in the JulyAugust 2006 Timelines column, “Death of a Sugar Daddy: The Mystery of the AFIPS Orphans.” Let’s start at the very beginning, with a description that may sound familiar to software developers.
“We may now state formally the purpose of this meeting: namely, to assess the adequacy of the designs of present working high-speed digital computers in order to point out the direction in which computer design should go, to make computers best for the jobs that they have been doing and for the jobs that they will have to do. This is basically an engineering or design objective, but it is clear that it also involves the users in an important way. This is a meeting of both builders and users, all of whom are actively interested in the field….
Let me point out four phases of life in this field through which you have all passed, more or less together. The phases were preceded by a period of pioneering by a relatively small number of people…. The first phase I have called the future, or building, or talking phase. The common denominator was a remark somewhat like this: ‘What a wonderful critter our computer is going to be.’ The second phase is what might be called the subjunctive, or debugging, or possibly the silent phase. The remarks that people made in this phase were, ‘It sure would be nice if we could get this thing to work.’ The third phase is the present, or working, or bright-look phase, and the remarks that people made at this time were, ‘Our computer is working now, but we haven’t had enough experience to judge it properly.’ The fourth I have called the past or getting-results phase. This again is a talking phase, and people tend to say nowadays, ‘It has been working fine; we are glad we built it, but we wish we had done this and this instead of what we actually did. However, we are going to fix that in our new model, and besides we are going to make it a lot simpler and more reliable.’
It is because of this latter phase that we have scheduled this meeting now.”
What meeting began with these comments about computers and their users? I omitted part of one sentence:
“The phases were preceded by a period of pioneering, in which a relatively small number of people developed computers such as the ENIAC and the MARK I calculator.”
The speaker was W.H. MacWilliams of Bell Labs. The year was 1951. The Joint AIEE-IRE Conference, with formal ACM participation, was organized by the associations of electrical engineers (AIEE) and radio engineers (IRE) following a successful 1950 meeting to discuss electron tubes, later called vacuum tubes.
The three associations continued to sponsor joint computer conferences. In 1961 they banded together to form a single parent organization, the American Federation of Information Processing Societies (AFIPS), and in 1963 AIEE and IRE merged to form IEEE.
User experience was a central concern. The 1951 conference may have been the first computer conference with published proceedings. Subsequent joint conferences met twice most years from 1953 to 1972, then annually as the National Computer Conference (NCC) from 1973 to 1988. After that, we gathered in specialized conferences to describe our wonderful critters, what we wished we had done differently, and how “we are going to fix that in our new model.”
Researchers of that era focused on journal publication, but papers contributed by major figures in computer science were published at these “must attend” conferences. This included oft-cited work by people who inspired and contributed to human-computer interaction. In our rapidly evolving field, one need not study history, but to get a sense of where current trajectories might take us, it can be useful to know how past trajectories ended, and what has remained constant over time.
An Orphan Finds a Home
AFIPS was large and prosperous. So much so that it built multimillion-dollar headquarters. Then AFIPS collapsed. In 1990, its records and collections were hauled to a dump. Four years ago, I asked “To whom, if anyone, did AFIPS grant copyright ownership of its books, magazines, and especially the proceedings of the joint and national computer conferences?” Neither ACM nor IEEE had a record of it. At one point near the end, AFIPS arranged with Springer to publish one of its journals, but Springer could find no pertinent records. A warehouse fire was mentioned. I tracked down AFIPS officers from that period, now in retirement; none had definitive recollections.
Why did I care? In the absence of a copyright owner, it could be risky to publish “orphaned” works that were influential in our early historywhat if “parents” appeared with an attorney at their side? Congress had considered facilitating the publication of orphaned works, but pulled back. Some might have significant commercial value, but conference papers from half a century ago were not among them.
The great news is that ACM has scanned an almost complete set of the proceedings, covering most of the years from 1951 to 1984. We are reconnected to a large piece of our history! The documents are available as digital library conference proceedings at http://portal.acm.org/proceedings/afips/. Not yet online are the AFIPS Office Automation Conference (OAC) proceedings from the 1980s, which include, for example, papers by Douglas Engelbart and one by Irene Greif describing the private symposium for which the term “computer supported cooperative work” was coined. (ACM plans to add the remaining volumes when accessible copies are located.)
A systematic survey of AFIPS contributions to computer science and HCI could be the basis for another article. With several volumes still missing, the proceedings include papers by a majority of the 55 Turing Awards winners. (Many of those who are not represented worked abroad.) This includes single-authored papers by 22 Turing Award winners, including Engelbart, Fred Brooks, Ed Feigenbaum, Alan Kay, Marvin Minsky, Allen Newell, Herb Simon, Ivan Sutherland, and others influential in HCI and related disciplines.
I conclude with two general observations prompted by these proceedings and the struggle to preserve them.
First, reviewing the past makes it clear that hardware research becomes obsolete relatively quickly in our fieldwe won’t learn much from papers on vacuum tubeswith software somewhat longer-lived, and process observations the most enduring. MacWilliams’s amusing description of optimistic self-assessments prompts reflection. The 1951 conference systematically brought in builders and users, so builders could learn from users and each other. Learning is slower when we assess our own handiwork. We emphasize the positive. Perhaps researchers should assess one another’s systems, or conferences should publish more negative outcomes. In my experience, frank assessments of unsuccessful system building get curt “you should have known better” rejectionsand do not enhance authors’ chances for renewed funding, either.
A dynamic industry that is growing overall can experience profound shifts that impact research and application. Digital technology extends its influence, technology-focused faculty and student populations expand, the demand for practitioners grows, and life seems good.
Second, consider the sudden demise of AFIPS and its conferences. The conferences juxtaposed research papers with expensive commercial exhibitions that could be very profitable. The NCC focused on mainframe computer exhibitors. When minicomputers became a big businessDigital Equipment Corporation became the world’s second-largest computer companyAFIPS organized office automation exhibitions and research conferences. However, AFIPS discouraged participation by PC and PC-software vendors. COMDEX advertised itself as “the NCC for the PC industry” and captured the exhibitors, while smaller ACM conferences (including CHI) captured much of the relevant research. In the 1980s, the mainframe companies and then the minicomputer companies foundered. NCC had a couple financially disastrous years, with large exhibit halls rented and not filled. AFIPS lost its large cash reserve, including money from the sale of its headquarters, and disbanded.
ACM and IEEE sought funds to replace their share of AFIPS revenues. Within ACM, focus shifted to Special Interest Groups and conferences, such as SIGGRAPH and CHI, which became a major source of revenue. This has been supplemented in the past decade by the Digital Library, which draws on conference proceedings for content.
What can we learn from this? A dynamic industry that is growing overall can experience profound shifts that impact research and application. Digital technology extends its influence, technology-focused faculty and student populations expand, the demand for practitioners grows, and life seems good. But AFIPS did not see clouds on the horizon. Could we experience equally rapid discontinuities?
Perhaps. The field has expanded tremendously, but ACM SIG membership has declined over 50 percent since 1990. Major conferences peaked in attendance years ago. In the May 2009 issue of Communications of the ACM, Ken Birman and Fred B. Schneider describe a crisis in computer science in terms of a “death spiral” attributed to systemic forces that tend to favor technically proficient but sterile research. They suggest that we command the ocean to roll back by returning to a journal orientation, but in our field the waves keep coming, each larger than the last.
We may avoid the AFIPS calamityfor one thing, our conferences are not tied to expensive commercial exhibitionsbut history tells us that change can be swift. It is sobering that when AFIPS collapsed, 40 years of research literature was suddenly inaccessible, and no one seemed to notice. We quickly moved on.
Young researchers and practitioners should expect and prepare for dramatic transformations in the field and in their careers.
Jonathan Grudin is a principal researcher in the Adaptive Systems and Interaction group at Microsoft Research.
©2010 ACM 1072-5220/10/0300 $10.00
Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee.
The Digital Library is published by the Association for Computing Machinery. Copyright © 2010 ACM, Inc.