LONG ACADEMIC AUTO/BIO|
(HIGHLIGHTING STUDENTS/MENTORS/COLLABORATORS/DEPT HISTORY):
Ronald Prescott Loui was a tenured Associate Professor of Computer Science and Engineering at Washington University for twenty years. He held guest appointments, at various times, in Philosophy, Linguistics, and Legal Studies, and was affiliated with Centers for Artificial Intelligence, Semantic Hardware, and Semantic Optimization (in Systems Science and Engineering).
Although originally majoring in Physics, Loui finished his degree at Harvard in Applied Mathematics (Decision and Control/Optimization), where he graduated in three years magna cum laude with a summa cum laude senior thesis that shared the ACM Forsythe Award. This thesis, ''Optimal Paths in Graphs with Stochastic or Multidimensional Weights,'' was published and translated into Japanese for republication. It continues to be cited and used as the basis for intellectual property (most recently in a patent by Harvard/MIT researchers). At the University of Rochester, Loui earned the M.Sc. in Computer Science and a Ph.D. in Computer Science and Philosophy. His degree was the first "Cognitive Science" doctorate at the U of R. His thesis advisor, the late Henry Kyburg, Jr., nominated Loui's dissertation for the Journal of Philosophy prize, where it was the unofficial runner-up. Kyburg's emphasis was where probability, logic, scientific reasoning, and epistemology meet, but Loui brought a greater interest in the social sciences to their discussions. Kyburg's pedigree bestowed on Loui a direct academic line to Leibniz and Kant.
This thesis, ''Theory and Computation of Uncertain Inference and Decision,'' contains a major chapter on the mathematics of argument using defeasible reasons. Over the next decade, this area would blossom and displace many of the more celebrated approaches to defeasible reasoning in artificial intelligence. Unlike the more famous work of his late friend and colleague, John Pollock, Loui's work emphasizes mathematical modeling, meta-theory, and historical connections to philosophy of language and law. Although default logic and semantic networks can be regarded as a mathematical logic of argument, Loui's is claimed to be the first direct attempt to mathematize dialectical debate as a logic of defeasible reasons. Loui corresponded with W.V.O. Quine, Nicholas Rescher, Herbert Simon, Isaac Levi, Roderick Chisholm, John Ladd, Carl Wellman, Peter Gardenfors, David Makinson, and Lotfi Zadeh on the historical precedents of the work, often crediting John Maynard Keynes and later H.L.A. Hart. Independent frameworks by Fangzhen Lin and Phan Min Dung appeared shortly thereafter, with more emphasis on semantics and reduction than derivation and procedure.
Loui was introduced to defeasible reasoning through the work of Donald Nute on d-PROLOG, but the mathematics was essentially Kyburg's. Loui took Kyburg's framework for probability and reused it on defeasible conditionals and belief revision (the idea was hatched during a trip to NYC when a Harvard classmate, Marialuisa Gallozzi, suggested he make better use of his typewriter). Kyburg had asked Loui to show that nonmonotonicity was really just a kind of moral certainty, but Loui found that the case for an irreducible defeasibility based in procedure was strong. David Israel, Jon Doyle, Thorne McCarty, and Rescher's monograph "Dialectics," would solidify this view.
At 25, Loui was a postdoctoral fellow at Stanford where he studied the foundations of measurement and quantum physics with Patrick Suppes, spent time at CS, CSLI, medical informatics, economics, economic engineering, SRI, Xerox Parc, and Rockwell Scientific Labs. He was funded by the Cognitive Science committee under Amos Tversky, invited by Paul Rosenbloom, signed under Yoav Shoham by Matt Ginsberg, and given space by David Israel. He actually spent most of his time at Tresider Union talking to close peers Eric Horvitz and Benjamin Grosof. And he began organizing his first academic meeting, the "Workshop on Multiple Inheritance and Defeasible Reasoning with Specificity."
At Washington University in St. Louis, starting at 26, Bill Ball asked him to help raise the intellectual reputation of the department. His first two doctoral students, Guillermo Simari and Gadi Pinkas, both had celebrated dissertations.
Simari's dissertation was completed a year after Loui's arrival, and it was nominated for the national ACM dissertation award. That award was given instead to Simari's fellow Argentine, Hector Geffner at UCLA under Judea Pearl. Pearl had recommended Loui for the Wash U job, and had encouraged Geffner and Loui's discussions. Geffner-Pearl's competing view of defeasible inference was at first more popular than Simari-Loui's (and it was probably the view Kyburg had wanted Loui to develop), but the Simari-Loui view in time produced a larger following. Family reasons caused Simari to return to Argentina where he had already been a mathematics professor, and he founded one of the most active research groups in computer science in South America.
Pinkas's dissertation was based on his independent ideas about Hopfield neural networks and fast constraint satisfaction. It drew the attention of Rina Dechter (UC Irvine) and David Touretzky (CMU) as de facto, external advisors. In 1992, the CS department found that it had two outstanding dissertations in one year, so a formal competition for the departmental nomination was held, and Pinkas won. The other dissertation student was recently Dean of the School of Computing at Georgia Tech and probably would have done very well in the national competition. Pinkas declined some academic offers and returned to AMDOCS, a large Israeli software company, where he was well-positioned when the company went public.
About 1992, Loui returned to his programming roots, briefly in LISP, but soon discovering the new gawk. Loui had met with Richard Stallman at MIT, while visiting former summer students Quinton Zondervan at MIT and Jennie Dorosh at Harvard. Stallman had spoken about the power of gawk. This began Loui's passion for scripting language programming that continues today. (Some say it reminds him of programming in Pick's DataBASIC under Gene Nathan Johnson in high school.)
Loui first used gawk to build a Harvard Alumni Internet Club, using mail forwarding on a fixed host like an FTP protocol (HTTP was being invented at the very same time). This was inspired in part by the observation that CalTech alumni, like Pete McCann, were allowed to retain their computer accounts so they could chat with each other online. The Harvard alumni internet club listed about 500 alumni, mostly in the computer industry or academia. Alumni could post static information about themselves as well as news items. The address, firstname.lastname@example.org actually forwarded email to the "hserver" running at ai.wustl.edu for a few years. David Israel and Lynn Stein were asked to lend their stanford.edu and mit.edu addresses for the recruiting letter. The site was more successful as a registry than a social network. It was closed when the World Wide Web idea of home pages began to catch on (also, the address ai.wustl.edu became allocated to an administrator's machine, and traffic was no longer forwarded to ai.cs.wustl.edu). By 1995, Loui figured that the Harvard alumni office would soon want to create its own social network, or at least its own repository of alumni home pages. Wash U student Seth Theriault briefly investigated selling or gifting Loui's programs to MIT for their alumni.
A short ACM SIGPLAN note at this time (invited by Ron Cytron) praising gawk for AI programming, is considered a "classic scripting paper." It was translated into a dozen languages for gnu distribution.
Defeasible reasoning was becoming popular in the niche research area, AI and Law. Continental doctoral students Henry Prakken, Tom Gordon, and Gerard Vreeswijk had noticed that defeasible reasoning could be used to describe European judicial argument. All produced theses that took long looks at Simari-Loui argument theory in the same year, and the theses of Arno Lodder and Bart Verheij soon followed (in addition to Simari's own students' theses in Argentina). There was more impetus from researchers like Giovanni Sartor (Italy), Katsumi Nitta (Japan), Guido Governatori (Australia), and Jaap Hage (Holland), all around the world. With inspiration from Edwina Rissland and her students, Loui soon claimed that common law analogical arguments from precedent also yielded to modeling with defeasible reasoning (the answer to the main concern that Cass Sunstein would later raise at a U Chicago symposium about AI and Law). During this time, Loui accepted invitations to give talks in Sweden, Holland, Germany, France, Italy, Japan, Argentina, and Spain, in addition to invitations within the US. Most of the conferences he attended during this period were at law schools (though at CMU, he was hosted by friends Bob Carpenter in linguistics, Katia Sycara in robotics, and Teddy Seidenfeld in philosophy, as there was no law school to visit). Bart Verheij and Fernando Tohme were Loui's official research associates during this time, but Prakken and Vreeswijk made short visits, and later, Carlos Chesnevar and Ana Maguitman visited from Simari's group.
The connection to law brought Loui into collaboration with graduating law student, Jeffery Norman. Norman had datamining ideas. He believed that the corpus of online judicial decisions could be mined to create a law-specific search engine upon which one could build a collaborative forum for legislative and legal argument (this would be an early kind of e-government community portal). By Fall 1994, traveling between Chicago and St. Louis, Jeff Norman had the idea that what one case said about another was more authoritative than what the case said about itself. Thus, pages were "harvested" for their citations of other cases (Loui preferred the term "harvest" since they were in the lower Midwest). The system, "Room5" (named after the address of the Supreme Court in Washington, DC), was considered for commercial development by the university. It could perhaps have become what FindLaw became. Norman wanted to compete with or be acquired by Lexis (and eventually, with Nexis on non-legal texts). The university consultant advised publishing the material on CD's, which the inventors rejected summarily. Norman believed in the dot-com business model, and Loui, who was obsessed with citation counts, detested the slow CD-based distribution of the Science Citation Index. Both wanted Room5 to be an online search service.
Gawk permitted rapid development of the site by students like Joe Altepeter (now a physicist) and Jessica Linsday (now a prosecutor). The search engine was hosted on interesting machines like the department's first dual-processor and Alpha-processor webservers (with initial hardware support from Michael Plezbert).
Loui made great use of undergraduate and even high school programmers on his innumerable web projects in this era. He liked to say that everything Time magazine claimed Patti Maes would be doing at the MIT Media Lab had already been done by his CS313 students. One of the biggest contributors at this time was Jonathan Pollock, whose simple regexp for removing HTML permitted easy screen scraping after web crawling.
The department had a small "Summer Undergraduate Research Assistantship" program that attempted to recruit talented non-Wash U students into its graduate programs. Taking over for Will Gillett, Loui considerably expanded this program with NSF funding. At its peak, the program was a national model for an REU program, with large numbers of underrepresented minority students from outstanding schools collaborating with Wash U peers and mentors on publishable work. Loui recruited at least a dozen Harvard students into this program. But it was the replacement of a Harvard no-show, Scott Hassan from Buffalo who replaced Steven Ketchpel, who set the bar for achievement. Hassan worked in Mark Frisse's medical informatics group with python mentor Steve Cousins. Both Hassan and Cousins found their way to the digital libraries group where google was first developed (it took considerable effort to convince Steve Cousins that the trip to Stanford would be worth severing his long ties to St. Louis, and Loui worked hard to bring Cousins back to St. Louis whenever possible, especially in leadership roles). Other students from this program have been on the faculties of USC, Maryland, Columbia, Johns Hopkins, Northwestern, and Yale. Some have joined research labs at MIT, Berkeley, and Google, and others have worked on major CG animation films like Kung Fu Panda and Matrix: Revolutions.
Loui's management of the REU program was notable for its research productivity and broad involvment despite a modest level of funding.
Loui had a superb undergraduate dissertation student, Diana Moore, who helped develop Loui's intuitions on negotiation. He had been using card games to model argument (e.g., with future Poker star and Berkeley math grad student, William Chen), and had moved to a similar, game-based modeling of negotiation (the late Anne Johnstone had suggested that negotiation was a nicer phenomenon than argumentation). To Loui, negotiation was a well-regulated dialogical process over a Nash-type payoff space that had dynamic aspects; it was something that a computer scientist could describe with modern ideas, but not a mathematical economist with 19th Century tools. Loui was in part reacting to the popular ecommerce ideas that Tuomas Sandholm had been producing at Wash U. Sandholm sought to import game-theoretic ideas into AI. Loui believed instead that AI should export its ideas to replace the game-theoretic conception of negotiation. Loui liked the ideas of computational linguists like Sandra Carberry and George Ferguson, but wanted a more familiar probability and utility framework. He was interested in explaining the phenomena described by Roger Fisher (whose course on negotiation Loui had taken as a freshman) and psychology-influenced authors in the management and law literatures. Moore's work with Loui presaged much of the award-winning work on argument and principled negotiation that was done in the multiagent systems community. Although Loui and Sandholm disagreed on the correctness of equilibrium as a "solution concept," Loui was committed during this time to helping Sandholm win the Computers and Thought Award, as Sandholm had been his best hiring decision since Scott Hassan (Sandholm was later "stolen" by CMU and there won the C&T award).
Many undergraduates contributed to Loui's "negotiation games" during this period, especially Harvard student Anne Jump (and many of the ideas with Moore and Jump remain unpublished). One of the ideas was contributed to Probability and Inference, a celebration of Henry Kyburg's career, in which a continuously varying expected value of settlement motivates unilateral concessions. A monograph, mainly written by Stephen Sachs (another Harvard student, with help from Winnie Yang of MIT) in 1997 on cgi programming in gawk, also remains unpublished. Perhaps the project was too commercial. Loui's activities as a Professor were always aimed at proselytizing rather than profiting.
After a brief collaboration with Marcel Waldvogel on speeding up hashing in virtual memory (that resulted in improved memory allocation in some gnu utilities), Loui met new faculty member John Lockwood. Lockwood claimed his FPGA's could do anything with network traffic. Loui suggested that they implement UNIX string processing tools, grep, sed, and the original awk in them. They recruited Michael Pachos (who had worked with Waldvogel and Loui on mmmalloc/independent-co-malloc) and James Moscola to write the gawk that would reprogram the FPGA's. Again, gawk development was fast. By Spring 2001 the device was ready for demonstration, and a patent application was written. Just after September 11, Lockwood was contacted by the Department of Defense asking about the range of capabilities of the devices he was producing. This began a seven-year collaboration with the intelligence agencies on devices that could do smart things with web traffic. The umbrella DARPA-TIA project was famously terminated, but special Congressional action continued the funding, and proper oversight, of this effort. In the final year, Loui was personally tasked with supporting some foreign language operations at one of the intelligence agencies. Although many publications resulted, many of the real discoveries remain unreported. The Lockwood-Loui project must be considered one of the most important and successful in the engineering school's history. Lockwood moved to Stanford after negative experiences dealing with the university over licensing.
Loui's final doctoral student was Moshe Looks, now at Google Research, who worked largely on his own ideas for speeding up genetic algorithms and genetic programs. His thesis was nominated for the IEEE dissertation award. Looks gave shape to Loui's last ideas in a pair of papers at JURIX-Brussels. They first showed a compositional language for discussing fairness in games. This was Loui's "holy grail" academic problem for a decade and a half: how to describe the procedural fairness properties of ordinary games, tournaments, and procedures. Then they produced an "academic joke" that depicted legislation as a never-ending attempt to project a high-dimensional objective function into too-few dimensions in response to "sharks" and "lambs" who were playing under existing rules too well, or too poorly. Loui liked to talk about "strategy extinction" and "acceptable ex-ante" or "derivable ex-post" asymmetries. A year later, Loui was a keynote speaker at JURIX-Paris, fulfilling a long-held academic desire to lecture at the Sorbonne.
In his final few years, Loui actually published more papers than at any time in his past, mainly because he had many co-authors putting his name on papers as a courtesy. Even now, his rate of citation on core contributions (defeasible reasoning, FPGA-grep, optimal paths, and modeling legal analogies) continues to rise. Loui published his final paper, heavily edited by IEEE Computer, "In praise of scripting languages: real programming pragmatism."
While working with Lockwood and supervising Looks, Loui taught more students than any other faculty member in the department. Because he insisted on co-staffing many of the CS100 labs personally and carried extra design classes, he often had 20+ classroom contact hours per week (more than the usual 3-6). Loui's most traditional class was Automata and Languages, where he could give proper mathematical definitions and proofs based on the Martin-text for a whole semester from memory. His signature classes were the AI programming classes, where entrepreneurial and research questions were addressed by ambitious design and programming teams. Loui would often compete directly with the undergraduates to design, for example, a better stratego player or simulated soccer player (Loui built his own soccer simulator in 1992, which pre-dated the AAAI soccer competitions by a few years). He was also proud of his baseball simulator, which he used to claim that Tony LaRussa was correct to bat his pitcher in the 8th spot (among Loui's other claims about optimizing batting order). One student review described him as the Haxxor professor who pwns the competition.
Loui had a track record of hiring the broadest range of teaching assistants, helping talented but wayward students complete their degrees (especially after tuition assistance would run out), and writing recommendation letters that "made the difference" with graduate and law school admissions. When Loui joined the faculty in 1988, there was a need for design experiences, thinking outside the box, and entrepreneurship. "Student discipline is the Associate Dean's problem" was one motto. By 2008, things had changed. Faculty assigned so many design projects that students were running out of creative hours. Suddenly, students needed disciplinarian professors, like old-timers Dan Kimura and Barry Kalman. But Loui found that the classes he was teaching (project-oriented, team-based or self-study, mostly created by department committees to increase enrollments, advertise subjects, or help students acquire skills without risk) made it nearly impossible to crack a new whip.
In 2007, Loui took a sabbatical to work informally and unofficially on the Obama for President campaign. Barack Obama and Ronald Loui graduated together at Punahou School in Honolulu, Hawaii in 1979, and Obama's memoir described his meeting Loui in fifth grade, in 1971, calling him "Frederick." Loui had invited Obama to be his banquet speaker at the ICAIL-2001 conference he organized in St. Louis. As Illinois State Senator, Obama had even joked that Loui's election to Speaker of the Engineering Faculty Senate was "a good start." Local TV reporter Betsey Bruce interviewed Professor Loui about his connections to Obama, and her piece was picked up by many of the Iowa stations early in the campaign.
The sabbatical and campaign brought Loui to Bay Village, Ohio. There, he began consulting for Cycorp at the Cleveland Clinic, gaining exposure to electronic medical records and outcomes reporting. Loui moved to Austin to work on temporary contracts with Cycorp, trying to help legendary AI maverick Doug Lenat realize his aims.
Loui claims that his gawk skills make him one of the fastest data cleaners and data analysts wherever he goes. He also remains a fast web services prototyper, especially for intranet-based improved productivity. He maintains a small-footprint lake house on a bluff near Lake Erie, much of which he has renovated with his own hands. He has been talking to friends about urban Unix training centers, introductory lectures on gawk, egovernment and cyberwarfare, and a way to replace relational databases and SQL for most applications.
Three of Ronald P. Loui's brothers remain in academia: Michael C. Loui, a full professor in computer engineering and former Associate Dean of Graduate Studies at U Illinois Urbana, Warren R. Loui, a lecturer on legal practice at USC's School of Law, and William S. Loui, a clinical oncology faculty member at the University of Hawaii Medical School. A fourth brother, Terrence D. Loui, held DoD computer admin positions in Hawaii before succumbing to childhood medical issues. The Loui brothers also have academically-published first cousins Joseph O'Connor, a UCSD full professor of chemistry and biochemistry and AAAS Fellow, Cassian Yee, a prolific cancer researcher in Canada, and Daryl H. Chinn, an academic radiologist at UC schools, second cousin Franklin T. Luk, former chairman of RPI computer science (who left for positions in Australia and Hong Kong), uncles Allen J. Chinn, contributor to the atomic bomb and SR71 Projects at Berkeley and Lockheed, and Leland J. Chinn, a pioneer in steroid chemistry. Mother Florence J. Chinn, published on medicare policy and preventing domestic violence.
Loui resigned from the Wash U faculty on June 1, 2007. These pages are hosted through the generosity of his former colleagues in the computer science department.