User Reviews & Comments: A Battle Between Engagement & Trust

Let’s have a look at best design practices for integrating user reviews and comments into your website.

There are many online services that seek users’ opinions to help other visitors make decisions. Sometimes this is a simple quantitative rating, as seen on Rotten Tomatoes with their average audience scores, or on Open Rice here in Hong Kong. It could be as simple as the ‘thumbs up’ or ‘thumbs down’ on YouTube or the ‘like’ button on Facebook. Other times user prose is solicited in the form of comment, forum replies, or reviews. Sometimes the user opinion extends to the judgement of other people’s comments. Witness YouTube comments being hidden for having too many negative marks, or the ordering of answers on Yahoo Answers or Quora by ‘most upvoted’. 

There is undoubtedly a value to user interaction. TripAdvisor is built on the concept, and a quick look at YouTube ‘likes’ is a great indicator of video quality or relevance. But cheating the system is often possible, and a discerning interweb-surfer will remember to proceed with caution and truckloads of salt.

How do we create trust in the user-generated content on our websites?

Good question.

Turtle Media was recently asked to create a user-comments system on a review website. The idea being that readers could supplement the main reviews with their own experiences and opinions on the topic. But the client was concerned; how to stop interested parties from blatant self-promotion, or conversely how to temper the vitriol and spleen from freshly angered patrons. It has been observed that reviews tend towards the negative, since negative experiences are more motivating to anonymous netizens than are average or positive experiences. Our options in this regards were several, and the decisions and considerations carry over to many situations on the web. So we discuss them here.

The simplest comments system is to present a comment form that invites some details of the commenter, but does not validate any of the fields.

In this case the user can claim to be CY Leung, with a prestigious White House email address should she so fancy. This may encourage more engagement since the form is so simple to complete. A time-poor visitor can make her voice heard in seconds. But how are we in turn to trust her? She may have never even visited planet Earth never mind used the product, service, or establishment in question. She might be the owner of the shop. Or the owner of the competing shop opposite. She might just be a nemesis of the cashier. A comment form that relies on honesty is very hard to trust.

Another way is to demand sign-up from a visitor. This has multiple effects. It gives an opportunity to validate the email address of the potential commenter. In the sign-up process the user must prove that they have access to this email address by clicking a link that is sent to them. They may be more restrained and thoughtful in their comments also, since they risk their reputation and their membership to the site. It also slows down the drive-by flamer who may have either calmed down by the time they are able to comment or slink away in frustration by step 2.

But both of these effects have their flaws. In the former it would be quite simple to use a secondary email address. Some serial ranters probably have several for this purpose. Worse, there are companies that will manage businesses’ reputations via just such endeavours, being paid to write favourable reviews (or worse still, unfavourable ones of competitors). In the latter the obstacle to quick commenting might also reduce the engagement in your forum. Fewer comments = less interaction = a less successful forum. We don’t want that.

Comment Validation Using Social Network profiles

There is another way. Social networks such as Facebook and Twitter offer another way to validate one’s identity instead of using the email address. For example using Facebook Connect a potential reviewer or commenter is presented with a pop-up asking her to allow the site access to their profile, thereby validating the user’s identity. If the user is already signed-in to Facebook this is a one-click process. Else she is asked to log in. This reduces the barrier to engagement presented by a multi-step process necessitated by validating an email address. But it also massively reduces the opportunity for abuse of the system.

Email addresses can be easily faked. They allow no methods for validation that a human might want to conduct. A second Facebook profile can be set up, but it is considerably harder. Even with a second profile, encouraging people to add you as a friend is even more difficult. So, when looking at a review or comment from a user that has validated with a Facebook account, you are usually (depending on privacy settings) able to tell that the identity is a valid, thriving account. There will be a recent image, recent posts, and a network of friends there. This is a very important element of the strategy for Airbnb. Before a host or guest trusts the other, they might ask if they can briefly access their Facebook or Twitter account via becoming friends. It’s virtually impossible for some malicious entity to create a Facebook profile with hundreds of connections. It’s certainly a greater effort than many would go to to rant about the local sushi train.

Final comment

In application, the best solution currently might be a mix of strategies. Not everyone is comfortable with social network identity validation, and for some web properties, relatively anonymous input mixed in might not cause much loss of integrity. In time we may see dedicated identification services, although efforts like OpenID have not caught on as fast as one might hope. We have services designed specifically for comments forms, such as DISQUS and IntenseDebate that will allow multiple methods of identification on the same forum.

We are already seeing the benefits of identity validation in larger ecosystems: Amazon, Android Store, iTunes, YouTube. People will become increasingly aware of the power of identity validation, and also more demanding of the web to provide user-feedback that can be trusted.