By Stephen Baker.

Houghton Mifflin. 256 pp. $26.

Reviewed by John Allen Paulos


ig Brother in a Chip

might be an alternative title for Stephen Baker's book

The Numerati

. With gee-whiz enthusiasm, the BusinessWeek columnist tells a number of fascinating stories about ever-cheaper, ever-more-powerful computer chips and the tools and techniques they make possible, tools and techniques that will increasingly and dramatically affect nearly every area of our lives.

The bulk of the book is devoted to examples. A particularly nice one concerns the chips and new software that will monitor our buying habits in supermarkets via the use of store cards and smart carts. These will note what we buy and infer from our purchases whether we're on a budget (relatively constant expenditures), following a diet (low-fat foods) or have fallen off one (high-fat ice creams).

The cart will also remind us if it finds we've forgotten something, and determine how brand-loyal and price-sensitive we are. It will discover our buying personalities and likely demographic characteristics, track our path through the store, and suggest changes in layout to stimulate sales of high-profit items.

Through interviews with scientists at various software companies ranging from giants like Yahoo and Google to a number of smaller boutique firms, Baker attempts to humanize the story of those he terms the "numerati." They are the mathematicians, computer scientists and others who are every day devising better software models of us as consumers, workers, patients, lovers, voters and even terrorists. Despite its topic, the book contains no mathematics, although Baker does periodically hint at notions from statistics, operations research, and graph and network theory that make the new software possible.

Interestingly, a few numerati are even analyzing blogs, because bloggers provide unfiltered, raw, generally honest reactions to products (from diarrhea medicines to golf clubs) that information-hungry companies want. Countless blogs are scanned for mention of these products (or issues), and the computer is taught to determine the sex, approximate age, and other demographic characteristics of the bloggers. The information obtained helps companies discern tastes and target ads (much as Google and Amazon are doing already).

On the drawing boards, too, are new medical software programs and devices that churn through volumes of information to monitor patients' health, especially that of older people. One example is a floor sensor that can detect changes in the speed and symmetry of Grandpa's gait, the frequency of his visits to the refrigerator or bathroom, and other characteristics. The sensor then determines whether he's especially unsteady in the morning (maybe too many meds the night before), or whether he's on the verge of a stroke, as well as the probability of a host of other conditions.

Phone software can detect if the time it takes Grandpa to recognize loved ones' voices has grown a fraction of a second longer and draw tentative conclusions and an invitation for specific tests. This would be especially helpful with Parkinson's disease, which often signals itself by changes in voice and gait as much as a decade before it's diagnosed by doctors.

The chapter on terrorism tells of the torrent of data available from the FBI, CIA, and publicly available databases, as well as monitored phone calls (some legally, many not), Web site visits, transportation records, tax filings, ubiquitous cameras, and who knows what else? This informational tsunami rushes past the National Security Agency's numerati, who attempt to pan nuggets out of the torrent.

Baker duly notes the danger of false positives in this endeavor. With ever more pieces of information and ever more superficially suspicious interconnections among them, the vast majority of the people picked out will likely be innocent. Note the million or so names on the airline watch list - and appreciate the continuing importance of habeas corpus.

Since mathematics is a somewhat imperialistic discipline, the same software tools and data-mining techniques useful in one domain apply in many other disparate realms as well. The ideas employed to find terrorists' messages and their senders are also used to find e-mail spam and its origins, to find dangerous molecules in our blood, to find suitable fits between workers and jobs, or to find potentially compatible mates. The connection among these various tasks is often nominal, consisting in no more than the fact that they all use similar mathematical tools.

Despite the slightly sinister sound to the title word


, Baker writes near the end of the book that these folks and their mathematical tricks will, for the most part, make our lives easier (unless we're trying to hide our hangover from the snooping floor). Far from being controlling, they will allow us to be more fully who we choose to be. Just as we all drive cars without understanding what a carburetor does, we'll all use these software tools without understanding how dynamic optimization works.