HOW MANY PEOPLE watch each program on television? For the last 30 or so years, the three networks and the major advertising agencies have been paying the A. C. Nielsen Co. in the tens of millions of dollars annually for the answers. Nielsen gets them by placing black boxes on all the TV sets in 1,700 randomly selected households, all of whose members are sworn to strict secrecy; the boxes tell Nielsen what station is tuned in on each set. From those responses Nielsen publishes the national program ratings. On those ratings hang the fates of "Falcon Crest" and "Miami Vice," of "Dallas" and "Dynasty" and "Hill Street Blues." An audience share of 30 or so keeps you on the air and gets you rich. A share a few points lower and you are cancelled and seeking other lines of work.

Now someone -- first a British-based company called AGB Research, then Nielsen itself -- is perfecting a better mousetrap, and the ratings business may be permanently changed. Methodologically the best you can say for Nielsen's procedures is the case Winston Churchill made for democracy: it is not a perfect system, but every other one is worse. It is not Nielsen's sample size or selection that is the problem; 1,700 respondents are, believe it or not, enough to permit statistically valid results within acceptable limits, and it appears that the households are chosen and their anonymity protected with the scrupulousness you would expect of a company whose $100 million business depends on its integrity.

The longstanding problem has been that no one can be sure who, if anyone, is really watching each set, whether these householders like and pay close attention to the program, whether they switch channels constantly and whether they watch the commercials. "The industry," a spokesman for AGB says, "wants a people-meter service." Networks and advertisers want to know who is really watching what. So AGB and Nielsen are both experimenting with a device with eight number buttons. Each household member would have a computerized "people-meter," coded for sex and age, and each would be asked to punch in when he began watching a program and punch out when he left the room.

You can see some problems here, too, but obviously the devices would provide more reliable information about actual viewing habits than Nielsen can now. But, we are sorry to say, even this is not the last word. It is a "first step," as Nielsen puts it, "toward totally passive people-metering." That means that researchers want to invent a heat or sonar sensor system that would determine what randomly selected household members are watching -- without the household members even knowing they were being monitored. That sounds perfect as survey research and perfectly awful in every other respect. But that is where the commercial appetites of advertisers and the technological capacity of business may be leading: there are tens of millions of dollars to be made. Are Americans ready to be passively people-metered? We actively hope -- and believe -- they are not.