Why online discussion forums often stray from original query

Ann Schlosser

Ann Schlosser

Good conversation sometimes takes a few detours.

But straying off topic is far less appreciated in conversation initiated to seek advice. On product purchases. Vacation rentals. Home repairs. Anything.

Yet diversion is precisely the fate of many online consumer discussion threads, according to new research by Ann Schlosser, a professor of marketing at the University of Washington Foster School of Business.

Schlosser finds that the content of advice offered in these online forums tends to be driven not by the original query, but rather by its first respondents. And when these early “experts” don’t really answer the question, they also steer subsequent respondents away from what was asked in the first place.

This meandering can lead to incomplete, irrelevant or even faulty recommendations.

“When posting a query to an online discussion forum, people tend to set out certain parameters,” says Schlosser, an Evert McCabe Endowed Faculty Fellow. “But respondents don’t necessarily keep those in mind when giving advice—especially in a group setting. This is because the conversation is evolving among the group members, who are each trying to affiliate more with their fellow experts than with the person seeking advice.”

Original social media

An estimated 20 percent of the population participates in one or more of the millions of Internet discussion forums hosted by the likes of Yelp, Google and Yahoo!, topical product sites such as Kelly Blue Book, and retailers such as Home Depot.

These forums are intended as a space to seek the learned wisdom of an online community of experts on topics that span the range of human experience: What’s the best hybrid car? How do I remove mold from lathe-and-plaster walls? Where should I stay near Disneyland?

A thread of responses typically follows, creating a vibrant exchange of information—at least in theory. But there are caveats. Responses in the thread don’t always address the particular criteria of the question, which is often quite specific.

Consider the example of a call for hotel recommendations near Disneyland. Say, a family with a 5-year-old and a 3-year-old will be visiting the park for five days in November, without a car, and is looking for an inexpensive but quiet hotel that is close enough to the park entrance to accommodate frequent naps.

The first response comes from a Disneyland veteran who loves, loves, loves the Buena Vista Lodge because it’s got such a fantastic outdoor water park and it’s only a 20-minute shuttle ride away from the park entrance. It costs a bit more but is totally worth it. Oh, and there’s free parking!

The problem? This enthusiastic response, while providing a Disneyland-area recommendation, does not actually address any of the specific parameters of the query.

And what if subsequent posters followed suit?

The (misguided) wisdom of crowds

Schlosser and co-authors Rebecca Harrison and Yu-Jen Chen set out to answer this question. They analyzed three actual discussion forum threads and created four more in laboratory settings in which participants were asked to play experts and post their opinion to the forum. Each was a variation on the same theme: a specified online query followed by an initial response that strayed from those specifications.

The results were consistent across the board. Forum participants followed the lead of the early respondents—even if that meant ignoring the specifications of the query that initiated the discussion.

Schlosser calls it an instance where the “wisdom” of crowds is misguided.

“You’d expect that people offering responses in an online forum would want to provide the most helpful content given the specifics of the query,” she says. “Instead, we found that people were less likely to mention a critical attribute if earlier respondents also ignored it—evidence against the ‘wisdom of crowds.’ In other words, we found that people would follow the herd, even when it’s taking them in the wrong direction.”

The need to affiliate

The study also expands upon a body of research on “hidden profiles,” which suggests that members of a group tend not to share information if they believe they are the only one who has it. Instead, they prefer to share only information that is not unique—even if it leads to suboptimal decision making.

This behavior is driven by the prevailing human desire to identify and affiliate with others in a group. Schlosser’s study demonstrates that this desire extends to “virtual” groups as well.

But in online discussion communities, the desire to affiliate is not with the person seeking advice, but rather with fellow respondents. The experts.

“Members may join forums seeking information, but they are more likely to continue participating when they affiliate with other forum members,” Schlosser says. “This desire to affiliate with others can systematically affect the information contributed by participants in online discussions, sometimes in undesirable ways.”

Redesigning discussion forums

It would seem that this tendency for online discussions to deliver incomplete or inaccurate recommendations could be corrected by a simple design tweak: allow participants to see only the query, not the prior responses.

While this design might produce more accurate and complete recommendations, Schlosser points out that it also would effectively remove the “social” aspect of this social media. And that bonding is the primary reason why people participate in these forums—to rub elbows with fellow experts.

To preserve this key motivator for participation, Schlosser suggests several more moderate measures. Online discussion sites could require users to outline clear parameters in their queries and encourage them to acknowledge respondents who address their query accurately.

One of the studies suggested an even simpler solution: remind respondents to answer the entire question. “When we prompted people to be more accurate in their responses,” Schlosser says, “they were less likely to stray from the question, even when previous respondents did.”

She adds that a larger message from the study’s findings is that individuals and businesses should perhaps read the wisdom of crowds more skeptically.

“Online discussions can be systematically biased in favor of repeating previously shared information,” Schlosser says, “regardless of its importance.”

Who’s Driving This Conversation? Systematic Biases in the Content of Online Consumer Discussions,” is published in the August 2017 issue of the Journal of Marketing Research.