����AV

Online technology has made real-time performance feedback a workplace reality. But a pair of Mason professors have found out about a major bias in the system. 

Online technology is fundamentally reshaping employee evaluations. In the last decade or so, companies such as , , and have adopted performance feedback apps that allow employees to "review" one another in real time. These apps take the 360-degree paradigm to its logical extreme by removing temporal, hierarchical, and geographical barriers to feedback. 

According to , assistant professor of information systems and operations management (ISOM) at , “People are trying to tap into new sources of employee engagement across all management and employment tiers. For millennials and Gen Z, instant communication is the norm, and they are not going to wait a year to get feedback. They want to know how they perform here and now, and be able to comment on other people’s performance in the same way. So when we merge these trends of social connectedness, instant communication, and use of technology, we come up with this wonderful application.” 

As with any breakthrough technology, though, appropriate use of real-time performance feedback depends upon understanding its inherent limitations. After all, increasing the scale and speed of feedback is not guaranteed to erase based upon gender, race, hierarchical position, etc. In a recent paper for , Petryk and her ISOM colleague , concentrate on a relatively neglected—but, as it turns out, subtly powerful—category of bias related to how individuals are embedded within the informal (i.e. social) network of the organization. 

Their co-authors were Michael Rivera and Subodha Kumar of Temple University, and Liangfei Qiu of University of Florida. 

Working from a unique data-set from technology provider DevelapMe, comprising nearly 4,000 instances of real-time performance feedback spanning five organizations, the researchers mapped the informal networks of each organization. They then compared reviews submitted by employees who were positionally embedded—i.e. those who moved in influential circles, though they may not themselves have been high-ranking—to ones by those who were structurally embedded, meaning they had larger clusters of weak ties. 

For example, both a C-level executive and their assistant could be considered positionally embedded. A middle manager whose work touches multiple teams would be structurally embedded. 

The professors found that positionally embedded employees tended to give higher scores to colleagues, while structurally embedded employees skewed negative in their ratings. 

Bhattacharya says, “Informal network bias could be explained as a matter of perspective. From atop the hierarchy, it’s difficult to see how projects came together and who made what happen. Positionally embedded people have a coarse rather than a granular view. Therefore, they may give highly visible individuals more credit than they deserve for collaborative work—for example, they may wrongly assume that a team member chosen to present a project to them was primarily responsible for said project. 

By contrast, structurally embedded employees have wider and more diverse networks and thus a much broader base of comparison. This makes them prone to detect and emphasize the flaws of co-workers.” 

Organizations using feedback apps such as DevelapMe can usually set limits on the number of anonymous reviews they allow, although the identities of anonymous raters are always visible to HR and senior leadership. The researchers found that anonymity magnified informal network bias for both structurally embedded and positionally embedded employees.  

Further, since reviews included both numerical score and explanatory text, the researchers analyzed how informal network bias influenced the wording of reviews. They saw that positionally embedded raters, though more generous with their numerical rating, were relatively neutral and formal in their written feedback. Structurally embedded reviews exhibited the opposite pattern: Comparatively strict in their scoring, but positive and encouraging in their written content. The researchers speculate this points to contrasting motives—constructive vs. motivational—the two groups had for delivering feedback. 

Easy ways to counter informal network bias, then, would be for organizations to carefully consider the amount of anonymity to permit, and for them to recommend or even require that each instance of feedback be accompanied by text. 

Beyond that, Bhattacharya and Petryk suggest that companies employ a combination of training and technological refinements to help address informal network bias. For example, positionally embedded managers should be reminded to temper their reviews with a bit more objectivity—perhaps peering outside their bubble to get a more complete picture of an employee’s work. Tech providers and consultants could use tools such as social network mapping to help organizations better account for informal network bias in their employee performance data. 

Petryk says, “Our data and findings show the mechanisms of how people—not necessarily high-ranking people—can have power over rewards, because at the end of the day the ratings will be factored into a formal evaluation. And bonuses will be distributed on the basis of the evaluation. How that decision is being made can be greatly impacted by the data that we analyze, and that we obtain from this network.”