Should performance reviews be crowdsourced?

October 8, 2013, 4:22 PM UTC

FORTUNE — Whether your company calls it 360-degree feedback, multi-source feedback, or crowdsourced performance reviews, chances are you’ve been required to fill out a form designed to let you praise or pillory your peers and the people above you (as well as your subordinates, if you have any). The practice, which originated with the German military during World War II, spread through American companies in the 1990s. It’s since been made easier by technology, with online questionnaires cutting out the cumbersome paperwork that used to be involved.

A 2012 bestseller, The Crowdsourced Performance Review: How to Use the Power of Social Recognition to Transform Employee Performance, added fuel to the fire. Author Eric Mosley is CEO of human resources consulting firm Globoforce, which counts Intuit (INTU), Amgen (AMGN), KPMG, and Procter & Gamble (PG) among its many big clients.

Extolling the virtues of 360-degree feedback in an influential Harvard Business Review blog post last year, Mosley pointed to the “wisdom of crowds” and declared, “Recognition is something that comes naturally to employees — they want to recognize their peers for great work” and are a font of “timely, measurable insights.”

MORE: Steve Ballmer bids farewell to shareholders

Maybe so. Experts estimate that between one-third and one-half of U.S. companies now use some form of crowdsourced performance appraisals — despite reams of research over the years that cast serious doubt on whether letting employees rate each other is fair, accurate, or useful.

One study, for instance, noted four primary reasons why crowdsourced reviews tend to be unreliable. One of them was that people “care more about the rewards associated with finishing the task than the actual content of the evaluation itself.” In other words, faced with one of those 360-degree forms to fill out, most employees just want to get it over with.

The biggest problem with crowdsourcing performance appraisals, however, is that the practice can muddy the legal waters if an employee sues the company.

“Employees’ performance reviews are an employer’s first line of defense against discrimination claims,” says Nesheba Kittling, an attorney at labor law firm Fisher & Phillips. Detailed documentation of job performance “provides support for an employer’s contention that it had legitimate, non-discriminatory reasons” for, say, a firing, a demotion, or a smaller-than-average bonus payout.

By contrast, Kittling says, “Peer-to-peer reviews, especially in a social networking environment, will likely distort the truth.” For one thing, employees often don’t have as clear an understanding of other people’s duties, their performance goals, or their success at meeting those goals as bosses have. Lacking enough information to judge colleagues’ work accurately, people tend to turn evaluations into “a popularity contest,” Kittling says. “They may be giving bad reviews to coworkers they don’t like, and inflating the ‘grades’ of those they do.”

The issue came up in a widely-publicized 2010 sex discrimination class-action suit, Chen-Oster v. Goldman Sachs. The complaint alleged Goldman’s (GS) crowdsourced review system “permitted unacceptable levels of subjectivity and bias,” which the plaintiffs said led to men repeatedly voting their friends into the top quartile of performance — where the juiciest bonuses go — and shutting women out.

MORE: Why China won’t sway the debt ceiling debate

Kittling believes much of crowdsourcing’s appeal arises from the fact that “managers don’t like giving performance reviews. They’re too busy, and they dread delivering bad news.”

But handing the task off to employees’ peers “just increases companies’ legal liability,” she adds. “It’s okay to solicit some feedback from coworkers — as long as there is a clear paper trail showing that any action the company takes is based on a supervisor’s assessment alone.” If you’re a manager, your appraisal of your subordinates’ work still has to be the only one that counts.