Very brief summaries of some of the technology.
— Data Mining, some methods very sophisticated, e.g. using neural networks
— Data collection, massage and correction, using various linearization methods, factor analysis, and its converse, synthesis, rotating the data as factorized into orthogonal components into frameworks suitable for application, not for scientific study, finding appropriate de-linearizations to undo the transformations which made the principal components or factor analysis into a linear algebra problem. Correction of data throughout this process using missing data algorithms, and considering answers to apparently redundant questions, some more prone to response-set effects.
— question generation by factor analysis on rows (questions) rather than columns (answers) when doing questionnaire data analysis, followed by searches through a large database of existing questions with some known response stats, to select questions which will get more and more of the information wanted, through an ongoing dialogue with the user — using a sharp instrument instead of a blunt one
— application of graph theory (or network analysis as sometimes — incorrectly — called) to do analysis of vertices (e.g. people) and edges (e.g. communcations channels or interpersonal relationships), with the aim of synthesizing more useful or “better” graphs, complete graphs, biconnected graphs, transitive graphs and perhaps especially, bipartite graphs. The most computation difficult problems arise in attempting to solve such problems, such as the notorious travelling salesman problem, which actually does arise, as e.g. when attempting to link stable spousal relationships through friendship relationships (man-woman -> woman-man -> man-woman -> woman-man, etc., forming a single closed loop where each person has exactly one spouse and one best friend.
— differentiating individual characteristics and the interpersonal relationships they imply with respect to time. Not just is this couple compatible today, but considering how each person is changing, will they grow closer or further apart, more or less compatible. Second derivative: is this changing over time. Perhaps they are indeed becoming less compatible with time, but rate of change is slowing, or perhaps reversing. Perhaps soon they will stop growing less compatible with time and may start growing more compatible with time.
— Group theoretic methods. To solve a Rubic’s cube, simple rotations to line up two cubelets will have the side effects of misaligning others. To solve the cube one must develop elaborate sequences of rotations, based on subgroups, which will do the desired move only. Similarly to optimize your own social network (never ever think of anyone like the government imposing some such optimization) — to work in your nearby social environment only, various sequences of changes can have the effect of accomplishing the desired simpler changes without the side effects of doing them directly.
— global social network optimizing, again by suggestion only, using simulated annealing algorithms where by the structure as a whole is treated as either a glassy or crystalline structure, as appropriate, with defects. These defects can be thought of as strained edges or as missing edges which should be connected in a graph (network). Relaxation methods can be used to model the possible annealing of this structure to remove these defects, and the resulting changes turned into suggestions for the user’s use, with tools to be provided.
Well, these are just sketches, absurdly brief, just hints of what can be done and how. I can talk endlessly about such things, and would be glad to do so for the possible amusement of selected technical experts from serious business interests, or well, just for mutual satisfaction, with Michael Ferguson’s erudite and refined polymaths. — dpw