DESPITE THESE OBJECTIONS, 'COMPUTER JUDGES' CAN OVERRULE THEMSELVES AND THUS HAVE THE POTENTIAL FOR ALLOWING LEGAL DOCTRINES TO GROW AND CHANGE. IN OTHER WORDS, ONE SIMPLE LINEAR REGRESSION MODEL MAY CORRECTLY DECIDE A SERIES OF CASES, AND IN THE PROCESS, OVERRULE A CASE CORRECTLY DECIDE A SERIES OF CASES, AND IN THE PROCESS, OVERRULE A CASE CORRECTLY DECIDED EARLIER IN THE SERIES. THIS IS ILLUSTRATED BY ASSUMING THAT ONLY ONE VARIABLE FACT IS RELEVANT TO A CASE DECISION, AND THAT FACT CAN OCCUR IN DIFFERENT INTENSITIES, MEASURED ON A SCALE OF 0 TO 10; FOR EXAMPLE, THE FACT MAY BE THAT THE POLICE USED FORCE ON THE DEFENDANT, AND THE INTENSITY MAY BE DIRECTLY MEASURED BY THE DEGREE OF FORCE USED. ZERO WOULD INDICATE NO FORCE BEING INVOLVED; 10 WOULD INDICATE NEAR-FATAL FORCE. FURTHER ASSUME THAT UNDER THE RULE OF LAW, 0 FACT INTENSITY OF POLICE FORCE RESULTS IN A DECISION FOR THE STATE, WHILE 1 OR MORE FACT INTENSITY RESULTS IN A DECISION FOR THE DEFENDANT. THE VARYING INTENSITY OF THE ONE FACTOR IN A NUMBER OF CASES CAN BE PLOTTED ON A GRAPH IN RELATION TO THE CASE OUTCOME. A REGRESSION LINE CAN THEN BE COMPUTED THAT WILL PREDICT CASE OUTCOMES ALONG A CONTINUUM OF INTENSITIES OF THE GIVEN FACTOR. A NEW REGRESSION LINE IS PLOTTED AFTER EACH CASE, SO THAT A MATHEMATICAL PATTERN FOR CASE LAW IS BUILT. AT EVERY POINT, THE REGRESSION ACCURATELY PREDICTS THE OUTCOME OF THE CASE BEFOR IT. YET, A CASE PREVIOUSLY DECIDED DOES NOT CONTROL HOW A SUBSEQUENT CASE PRESENTING THE SAME FACT WILL BE DECIDED. IN A SENSE, THE MATHEMATICAL PROCEDURE HAS LEARNED; CHANGING CIRCUMSTANCES MAY BE ACCOMMODATED TO GIVE DIFFERENT RESULTS. THE 'COMPUTER JUDGE' DOES NOT ALWAYS GIVE THE SAME JUDGMENT ON THE SAME FACTS IF 'CIRCUMSTANCES' HAVE SUFFICIENTLY CHANGED. FOOTNOTES AND DIAGRAMS ARE PROVIDED. (RCB)