So, what’s your rating? How the trust economy is taking over our lives10 min read . Updated: 08 Oct 2018, 03:26 PM IST
Ratings are becoming crucial in how we transact. They can also be manipulated. So why do we increasingly trust them?
Bengaluru: Uber ratings have been evoking joy and frustration for years. Though most tweets and posts about them are light-hearted, there are deeply anguished ones as well—a Reddit thread titled Uber Rider Rating going down (in spite of tipping, politeness, etc) is a good gateway drug to get into the messy conversations and debates around why drivers and riders rate each other the way they do.
One user, who claims to be a driver, lists over a 100 reasons why a driver may give a rider less than 5 stars. His list includes, among obvious stuff like throwing up or snorting drugs in the car, esoteric complaints as “Did you talk like a frat bro or sorority girl? Did you beatbox?" and, a favourite, “Did you talk about CrossFit?" On a different thread, a driver discloses his simple rule: “If I would be glad to repeat the ride, 5 stars. If not, 1 star, as a 1 star rating is the only way not to be matched with that rider again."
There are other signs as well that people take ratings seriously. On Tinder, it’s not uncommon to find “Rated 5 Stars on Uber" on profiles as a sort of validation of one’s respectability as a prospective date, and there have been accusations of racism and sexism when it comes to rating drivers and riders, respectively.
Consider this reporter’s experience: my Uber rating is a miserable 4.4, having climbed up from an even more miserable 4.37 after a recent trip to Delhi. When I revealed my rating to friends, I was greeted with incredulity, followed by the usual questions about whether I had thrown up in cars or been violent and abusive (I haven’t).
Slowly, a pattern emerged: most men I knew had perfect ratings of 5 or 4.8 at worst, while many women reported lower ratings of below 4.5. Among the people I informally polled in Bengaluru, the lowest ratings (below 4.4) belonged to women who did not speak to drivers in the local language and used Uber frequently (the more often you use the service the more rides it takes to change your average). I ticked all the boxes. This is conjecture, but I noticed that my rating climbed up by a few micro points after my trip to Delhi—possibly because I had better communication with the drivers and they gave me perfect 5 star ratings.
A matter of trust
The sharing economy depends on mutual accountability, and ratings are a huge part of that, but the system is also subjective and inaccurate—a system based largely on feelings and emotions rather than hard data. Ratings are essentially a system of building trust—and never before in human history has it been so ephemeral; so easily gained and lost.
It’s growing. The mutual rating system is visible on services like Uber, Ola and Airbnb, and an integral part of review and feedback fields on, say, an Amazon or TripAdvisor. Insiders say that many service-oriented start-ups in India, such as delivery and on-demand salon services, have discreetly added the facility for customers to be rated by personnel in order to weed out difficult users. The facility is not visible to the user but your food delivery guy and the lady who comes to do your eyebrows might be rating you even as you rate her.
Rachel Botsman, a lecturer at Oxford University’s Saïd Business School and the author of Who Can You Trust?, talks about how the distribution of trust via technology impacts behaviour. “I don’t always bother to hang my towels up when I’m finished in the hotel, but I would never do this as a guest on Airbnb. And the reason why I would never do this as a guest on Airbnb is because guests know that they’ll be rated by hosts, and that those ratings are likely to impact their ability to transact in the future. It’s a simple illustration of how online trust will change our behaviours in the real world, make us more accountable in ways we cannot yet even imagine."
In an anguished Medium post from a few months ago, Jackson Cunningham, a furniture designer from Canada, wrote about being locked out of his Airbnb account without an explanation. After several email exchanges, Cunningham got this response from the company: “Please understand that we are not obligated to provide an explanation for the action taken against your account. Additionally, we consider this matter closed and will no longer reply to any inquiries regarding your account."
“Guests receive a star rating out of 5 for cleanliness, communication and observation of house rules," said Amanpreet Bajaj, country manager, Airbnb India, in response to emailed questions shared by Mint. “So a low guest rating can be a result of damaged or misused property, miscommunication of booking dates or other guest information and defiance of the host’s house rules (which hosts are free to set as per their individual requirements)."
Most of us can easily recall a time before Uber and Airbnb, but that doesn’t mean the fear of being locked out of them is any less real. On 5 September, Sydney Morning Herald reported that Uber users in Australia with lower-than four-star ratings would not be able to use the app, earning a ban for six months. This was the cue for an immediate media meltdown. Headlines screamed that Uber was banning passengers for low ratings. Twitter was lit. And inevitably, the shadow of Black Mirror, the science fiction TV series, rose to engulf all discourse on the subject. Many quoted the episode Nosedive, which predicted a sinister fake-cheery future in which one’s social rating would grant or withhold privileges or even access to essential services.
However, reality is more nuanced. A high-ranking Uber executive in Australia said that getting an average rating of below 4 wouldn’t be “an accident", which essentially means that it would take more than one bad trip to push you into the blacklist. Uber also added that users wouldn’t be immediately booted out of the system—there would be warnings and a grace period, though the company did not specify how long this would last. “While there are no immediate plans to update our existing community guidelines in India, we will continue to work towards making Uber the best possible experience for both riders and driver partners," an Uber spokesperson said in response to an email from Mint.
Trust can be broken
Clearly, we cannot summarily dismiss the general apprehension that the “rating culture" will lead to us being excluded from the digital economy because of events beyond our control. And increasingly, the digital economy is important not just to our comfort and convenience but to our very existence as we link essential services, such as banking and payments, to identity numbers, apps and social media, and give access to our thoughts and ideas to political parties and their data teams to manipulate.
In a hyper-connected digital world, where your financial and social behaviours are so closely linked, social scores impacting your real-life worth and even identity is a legitimate fear. Today you may get locked out of your Twitter account, or be banned from posting reviews on Amazon, but tomorrow it might be your bank account.
Taken to the extreme, personal rating systems can be potentially life-threatening. Hitch, a social ride-sharing service run by China’s Didi Chuxing, allowed drivers and passengers to review and rate each other by appearance till May 2018, when a female customer was murdered. The service had earlier allowed users to comment on each other’s appearance, with some female passengers tagged as “goddess", “adorable" or “long legs", reported Reuters.
Conversely, mutual rating systems do provide a semblance of egalitarianism to the whole idea of service. Service providers were told ‘the customer is always right’ for centuries—a feudal mindset that is marked in India and other Asian countries with horrible structural imbalances of power. It is also prevalent even in more equal societies where the expectation is that the server will bend over backwards to please you, no matter how obnoxious you are to her. The old joke about the waiter who spits into your food behind your back needs to be rested—today the waiter can, and should, be able to puncture the swagger of an over-entitled guest.
Social credit ratings
As has been widely reported, the “social score" is already a reality in China. The social credit rating emerged from the need for credit scores to receive bank loans and credit cards for a society where the majority lacked credit and banking history 20 years ago. Wired magazine reported in December 2017 that the aim is for every Chinese citizen to have their own file on a searchable database, compiling data from public and private sources by 2020.
In India, too, fintech start-ups and non-banking financial companies (NBFCs) have started accumulating credit scores based on publicly available data, which may include social media ratings and other activity. Fintech start-up CASHe, which disburses short-term personal loans, has developed its own algorithm to determine the credit-worthiness of its customers, most of whom are millennials with little or no banking history.
“The current banking-based system of giving loans based on credit ratings is a dumb system," says CASHe founder V. Raman Kumar. “It is entirely dependent on former credit activity. We felt this had to be redefined."
The company has devised a ‘Social Loan Quotient’ (SLQ), a social behaviour-based alternate credit rating system that Raman says uses 15-200 data points to determine a score on a 1000. The CASHe app uses “rich phone data" to track user behaviour such as phone usage, LinkedIn and Facebook profiles, what kind of apps you have, how often you use them, whether you use e-commerce and financial apps etc. SMS gives access to spending history thanks to bank SMSs and OTPs required for online transactions. It tells them whether you have paid EMIs and SIPs, and if you defaulted on your Netflix subscription.
Besides this, they also look at your “cohorts"—for instance, people who work in the same company as you or are from the same educational institutions. Kumar says they are building the SLQ as an industry-wide score that may come to replace or supplement conventional credit scores like Cibil. “Earlier, loans were asset-driven. Banks wanted physical security. But today’s transactions are digital; there’s less focus on owning and more on renting. If banks are not looking at all this, they need to change their behaviour," says Raman.
When Botsman says reputation is going to become a currency more powerful than our credit scores in the real world, she’s not too far off the mark. But one of the flaws in ratings is systemic. Apart from concerns over privacy and ways in which biases can creep in, there’s the phenomenon of “reputation inflation".
In a March 2018 paper with the same title, researchers Apostolos Filippas, John J. Horton and Joseph Golden from New York University’s Stern School of Business argued that the effectiveness of post-transaction ratings in the shared economy deteriorates over time.
“The problem is that ratings are prone to inflation, with raters feeling pressure to leave ‘above average’ ratings, which in turn pushes the average higher. This pressure stems from raters’ desire to not harm the rated seller. As the potential to harm is what makes ratings effective, reputation systems, as currently designed, sow the seeds of their own irrelevance," says the paper.
The researchers note how on Uber and Lyft, it is widely known that anything less than 5 stars is considered “bad" feedback, and this leads to the conclusion that if feedback scores are rising, it could be because of two distinct—but not mutually exclusive—reasons: raters are becoming more satisfied, or raters are lowering their standards. “This second possibility—giving higher scores despite not being more satisfied— can be thought of as a kind of inflation. This ‘reputation inflation’ potentially makes feedback completely uninformative," say Filippas, Horton and Golden.
That’s why by making the ratings economy valid criteria for deciding financial and personal worth, we might be putting our trust in a fickle god.