The day after he became President Reagan's science adviser in 1981, George A. Keyworth II made it clear he was going to reverse the traditional role of the job. Instead of funneling the best ideas and advice of the nation's scientists to the president, Keyworth said, he would help shape White House policy on technical matters, then relay it to the scientists.

The most notable consequence of this role reversal, according to many scientists, is Keyworth's adoption and promotion of Reagan's Strategic Defense Initiative (SDI), an advocacy that comes despite deep divisions in the scientific community over the feasibility of building a shield against nuclear missiles.

Many scientists support the SDI, also known as "Star Wars," but many say the program is scientifically flawed. Opponents include more than half of the physics professors at 14 of the country's top science-oriented universities. They have signed petitions refusing to work on the SDI.

Many scientific leaders, including some former White House science advisers and others who have served previous administrations on the President's Science Advisory Committee, have expressed dismay that Keyworth has not only adopted the president's position but has become a vigorous public advocate of the SDI.

Critics contrast Keyworth's role with that of his predecessors.

During the Nixon administration, for example, science adviser Lee DuBridge, taking the prevailing view among scientists, opposed President Richard M. Nixon's plan for the Safeguard antiballistic-missile system. Nixon was so displeased by lack of support from his science advisers that he eventually abolished the position. President Gerald R. Ford restored it.

During most of the Johnson administration, Donald F. Hornig was science adviser, but he lost favor by opposing President Lyndon B. Johnson's plan to build a supersonic transport (SST) aircraft, a stance in agreement with the dominant scientific opinion.

By contrast, Keyworth is seen, as one of his predecessors put it, as a "spear-carrier" for the SDI who spends much of his time not advising the president but trying to persuade scientists to put aside their doubts.

Saying he spends about half of his time on the SDI, Keyworth dismisses the criticism, saying he has not behaved much differently from previous science advisers.

"Most of my predecessors . . . realized that they could not be effective simply as spokesmen for the scientific community," Keyworth said. "They understood that they served the president and had to behave as guests in another man's home."

Critics, including predecessors who asked not to be identified, insist that Keyworth goes too far in trying to change the minds of dissenting scientists.

"Some of his speeches have been a little strange," one former science adviser said of Keyworth. "He spends a lot of time lecturing universities, telling them they have to change and telling scientists they should stop criticizing the government. He makes you think it's unpatriotic to raise questions about SDI."

One former advisory committee member accuses Keyworth of "squandering the credibility of the science adviser" within the scientific community.

One scientist complained that his critical analysis of the SDI was not taken seriously by Keyworth's office. David Parnas, a top computer-programming expert, tried to give Reagan his written evaluation of a potential Star Wars system. Parnas had worked on military computing projects for many years and had been chosen by the Pentagon's SDI Organization (SDIO) to serve on its advisory panel on computing.

Parnas attended the first session, became frustrated by what he saw as the SDIO's lack of realism, resigned and then wrote eight technical papers examining what he contends are nearly insurmountable computing problems in attempting an antimissile defense.

"I thought it would be good for the White House to have these papers," Parnas said, "so I sent them to Keyworth. Two and a half months later, I get a reply from somebody in Keyworth's office.

"It says: Thank you for your papers. We appreciate the care you put into this. However, the president's directive has given us a mandate that many of us believe in and we are compelled to follow.

"What worries me," Parnas said, "is I see the government depriving itself of objective, informed advice on technical matters."

That lament is shared by many in the scientific world. Scientists tend to believe their opinions should carry extra weight because they are based on a rigorous analysis of objective evidence.

Keyworth contends that his advocacy of the SDI emerged from objective evidence. Originally, he said, he was skeptical about putting expensive laser battle stations in space because they would be vulnerable to Soviet attack. He said his mind was changed by the findings of an advisory committee in 1982, before Reagan proposed the program.

The panel, convened to examine prospects for modernizing the U.S. strategic military capability, noted a key advance that made it possible to keep the laser on the ground and relay its beam with relatively low-cost mirrors in orbit.

Keyworth said that finding contributed to Reagan's SDI decision. He also said his reliance on the panel shows that he was following his predecessors' practice. The panel report, however, is "highly classified," he said.

"I realize our work was not done with the same degree of openness as might have been done in the past," Keyworth said. "But I really don't think my support of SDI comes from as obvious a political conviction as much of the opposition to SDI has."

The position of science adviser was created by President Dwight D. Eisenhower in 1957 in the panic that followed the Soviet launch of the Sputnik satellite.

The president, it was felt then, needed ready access to the nation's best scientific minds and should have a prominent scientist in the White House who could tap specialists as needed.

Before Keyworth, each adviser was a prominent scientist with great respect among his colleagues.

Reagan reportedly was turned down by more than a dozen persons before settling on Keyworth, a virtual unknown beyond the Los Alamos National Laboratory. There he supervised physics research, including some on lasers and particle beams, both of which have been proposed as antimissile weapons.

Keyworth reportedly was endorsed by Edward Teller, the hydrogen-bomb scientist who befriended Reagan during his days as California governor. Teller, who has advocated antimissile defenses for more than 15 years, is thought to have played a major role in convincing Reagan, years before his 1983 "Star Wars" speech, that strategic defense was needed and possible.

Early in his tenure, Keyworth aligned himself with the president. "My set of principles are the same as his," Keyworth once said. "My policies are his policies, and his policies are my policies."

As a latecomer to the Reagan administration, he inherited a budget that cut science spending deeply. Scientists complained, but Keyworth branded them as arrogant and said the cuts were needed to weed out mediocre scientists.

Subsequently, however, he has been able to protect science budgets and even see them increased in an era of stringent cutbacks. Scientists generally credit Keyworth for this and recognize that his clout in the White House is based on his alliance with Reagan.

In 1982, Keyworth was Reagan's point man on the Dense Pack strategy for clustered basing of the MX missile. He was supposed to persuade opinion leaders, such as newspaper editorial-board members, that Dense Pack made scientific sense.

Dense Pack was one of many MX-basing ideas considered by Frank Press, science adviser to President Jimmy Carter. Instead of choosing the most politically appealing option and advocating it, Press convened a panel of experts who concluded that all basing methods had problems -- a verdict that has held up.

Dense Pack was eventually abandoned. MX missiles were planned for the same kind of underground silos declared dangerously vulnerable in the first place.

Keyworth said his panel meets regularly to assess SDI progress and is pleased with developments. Its deliberations are secret and no public reports are issued.

"Keep in mind," Keyworth said, "SDI is still only a research program. We're a long way from any decision to deploy anything.