Page:Sm all cc.pdf/126

 Ideally, each measurement given in a publication would be accompanied by a calculated estimate of its precision. Precision estimates, however, generally require replicate measurements, which may not be available. The use of significant digits may have to suffice. The number of significant digits is equal to the number of digits that are reliably known, ignoring leading zeros.

Although the rules concerning significant digits are simple, few of the current software packages honor them. Some follow the conservative approach of assuming that all digits are significant (e.g., 1÷3=0.333333...). Some strip off trailing zeros whether or not they are significant; for example, a series of numbers accurate to ±0.01 might appear as 1.14, 1.1, 1.07, and 1. Most maintain a user-selectable constant number of digits to the right of the decimal place. None of these conventions is appropriate for publication.

The word computer is no longer appropriate. The proportion of computer usage devoted to computation is steadily decreasing. Many of the recent computer developments have had little to do with computation. Of particular interest to scientists is the extent to which computer networking is revolutionizing information handling.

Efficient information handling has always been an essential aspect of scientific method. Even the early days of science had more observations -- mostly irrelevant -- than a mind could encompass; an example is Leonardo da Vinci’s quicksilver mind and notes. Today, information handling is a mantra of our technological society. Are we witnessing another transient enthusiasm, or are we truly challenged to adapt or be left behind?

Research faces two information problems -- locating and organizing relevant information. These problems are relatively minor in the course of one’s own experiments, although they certainly are felt while writing up results. The real hurdle is in dealing with the vast published literature. All memories are fallible, especially my own. Where was I?

The first step in information handling is skimming or digesting a scientific publication. These days, we usually have a personal copy of the paper rather than the library’s, so we are free to mark up the paper with underlines and marginal comments. To organize information from several papers, many people simply group and rescan a stack of reprints. Others prefer to take notes, either on a pad or laptop. A virtue of the latter is easy reorganization, because association is essential to pattern recognition. Furthermore, typing is faster than writing, and the Find command is a great time saver.

Ambitious schemes for information handling tend to fail. First the scientist falls behind in entering data into the system. Later, the backlog is so great that the system atrophies.

Is information handling by computers more efficient than by scientists? For straightforward sorting, bookkeeping, and information archiving, the answer is yes. The quantity, or content, of science is doubling every five years, so the need for efficient data handling is undoubted. Use of the Internet and the World Wide Web is growing exponentially, and every scientist faces the question of how much to employ these valuable tools. Both publications and published data are becoming more available on the Internet. We can anticipate, as a result, increased awareness of relevant publications and more analyses of data by individuals other than the one who collected the data. Where better to exploit the Information Age than in the quest for answers to scientific questions?

Whenever one develops a hypothesis, the first step is to see whether or not it survives the test of existing data. If we decide that none of the previous relevant experiments was appropriately designed to provide a diagnostic test of the hypothesis, only then do we conduct a new experiment. Scientific progress does not imply, however, that the same person who generates hypotheses tests them. Already, many scientists are tapping the information river to produce papers that present no