In an article in this week's Career Development Center, Robert Austin explains his theories on managing knowledge workers. Here he discusses how those theories might apply in a laboratory setting.

CDC: The typical academic laboratory employs "agents" with a wide range of expertise and experience, from the unmotivated undergraduate student who understands little about the work, to the experienced technician who is, in effect, the lab's Chief Operations Officer, to the advanced postdoc who may know more about the work than the "principal" and will be running her own laboratory in a year or two. It seems obvious that the "agency model" is more applicable to some employer-employee relationships than to others. Is it the same in the business world? Does a manager need to approach different principal-agent relationships in different ways?

Austin: Absolutely. You've summarized the gist of what needs to happen when people think about managing performance. Do the assumptions about motivation of employees and observability of performance formalized in agency models hold? Or don't they? The approach you take has to be contingent on the answers to these questions.

My focus on knowledge work has led me to explore areas where the assumptions don't hold very well. But in some cases (e.g., unmotivated undergraduates), laboratory agents may be excellent candidates for more traditional supervision. I should note, however, that people sometimes overestimate the extent to which the conventional assumptions apply. Often, for example, managers assume they are more qualified than they really are to judge their employees' performance.

CDC: So what are the implications of this? If a manager (say, the PI of an academic biomedical research lab) assumes that they're more qualified to judge their employees' performance than they really are, what management errors do they typically make? How does it affect the performance of the organization (e.g., the laboratory)? Is there any way to guard against these mistakes?

Austin: When a manager assumes signals of his or her employee's performance are better than they are, we get the classic pattern of dysfunctional performance measurement.

The manager relies on signals that he or she assumes are good results measures. In fact, the employee knows ways to make the signals look good that the manager hasn't thought of and that have nothing to do with results. As time passes, measures of performance trend upward, which makes the manager happy. Unbeknownst to the manager, however, the measures have less and less to do with actual performance. Eventually something has to give and the whole things blows up--you get an Enron or something like it.

The insidious thing here is that a manager can't tell the difference between a solid performance measurement system and a dysfunctional one, until the latter blows up. To a manager making unwarranted assumptions about his or her ability to monitor employees, a dysfunctional performance measurement system looks just fine--measures are increasing. But workers are spending less and less of their time on productive work, more on "gaming" the measures.

Another lousy feature of such systems is that they punish workers who have too much integrity to game the measures. When a dysfunctional system blows up, it's not uncommon for the honest worker who refused to game the system and who therefore wasn't doing as well on measured performance to catch a disproportionate amount of the blame. Organizations that rely on dysfunctional performance measures have a long-term chronic disease that ultimately drives out workers with integrity, in favor of those who are willing and able to game measures. Once again, think of Enron.

Because of the dire consequences of dysfunctional performance management, I usually advise people to make very conservative assumptions about their ability to monitor the work of subordinates. This holds doubly for knowledge-work settings.

By the way, over the next several years, I predict we'll learn these lessons about dysfunctional measurement all over again from the craze over testing in our public schools. This, too, seems to be a lesson from my research--for some reason, we don't learn very well in this area. Maybe it's because you can't convince people that they don't understand what their subordinates are doing. Maybe that's an inherently threatening thought.

CDC: You've suggested that measuring the performance of knowledge workers is very difficult, maybe impossible, but many people think it is also essential. How do we decide who to promote? Who to hire? Who to award a grant to? Are there any "conservative" performance metrics that we can turn to, or, perhaps, a decision-making strategy that doesn't rely on performance metrics? If we dispense with measurement-based management, what can we replace it with?

Austin: In my view, the key is getting the lines of communication between manager and employee open and keeping them that way. It's really difficult to impose monitoring on knowledge workers. That shuts down communication more often than not. Ironically, though, knowledge workers are often quite willing tell all about their work, good and bad, if they feel they're being treated fairly and professionally. There are still issues to be managed--for example, your overall assessment of an employee's performance may not match theirs (90% of drivers think their driving skills are above average, and the same often holds true for employee performance).

But if the lines of communication with your team are wide open, you'll know who the standout performers are, as will everyone else on the team. Promotions will be fair and viewed that way, and even if they're not there is a basis for discussing that--for employees to air their grievances. If an opportunistic worker tries to game or posture, open communication lines will eventually help you catch on to that--you'll notice that there are things the employee isn't telling you, or you may hear from other employees that something is amiss within the team. Being a good listener, hearing between the lines, is part of it. You have to really be interested in what your people are doing. Doing this well is work. You can't just sit in your office and read "performance metrics." Often the performance metric approach is a punt, a poor excuse for really managing.

Jim Austin is the editor of Science Careers. @SciCareerEditor on Twitter