On the need for a transitive property of trust, at scale
I'm having this exact problem right now trying to hire movers. Every moving company has some reviews that are like "they showed up hours late and smashed everything" and other reviews that are like "they were perfect, I would entrust them to carry the One Ring to Mordor."
And yet, as you point out, this problem seems pretty well solved for goods. I rely heavily on Amazon reviews and Wirecutter recommendations, and I can't think of a single time I've been disappointed. Why can't we solve it for services too? I think there may be three reasons.
One is that services are infinitely variable. Every Honda Accord is the same, so if you got a good one, I can get a good one, too. But every doctor's visit is different, so how can I know that your experience will be predictive of mine?
Another is that services are private. There's just no way for me to verify whether your accountant did a good job for you. I don't get to take them for a test drive, or sit on them at the store and see if they're comfortable.
But I think the most important one is that people's opinions about other humans are way more variable than their opinions about objects. We all want pretty much the same thing out of a laptop: runs fast, battery lasts long, doesn't crash. So I can trust you to vouch for a laptop. But people have all kinds of weird opinions about the way services should work. Some people expect lots of smiles and chat; other people find that grating and inauthentic. Some people think not returning a phone call within 24 hours is inexcusable; other people don't care. Some people see service workers as their own temporary serfs; other people see them as, you know, people. Add on top of this the fact that people are super weird themselves––maybe they're terrible at packing and that's why their stuff got destroyed, maybe they left out a bunch of documents and that's why their tax returns were wrong, maybe they were rude and that's why their waitress was too.
All that means it's hard to trust strangers about services. It's like asking a random person to pick someone for you to date. Anyway if you know any good moving companies that work in NYC and DC, let me know.
I’ve always thought that the problems are just unsolvable at scale unless a corporation that’s customer-obsessed steps in. Imagine the Amazon of insurance or movers, for example. They’re extremely concerned with their brand and will do root-cause analyses whenever a disgruntled customer complains to the CEO. And customer-obsession is baked into the culture to the point where it’s a factor in performance reviews. Not that they’re perfect, but small communities are also not perfect.
Great article as usual!
One of the problems seems to me to be lack of dimensionality in reviews. I'm working now with an artist whose style I love and who is very communicative, but has had to push back the deadline. The review system will favor a single numerical score that's supposed to encompass every aspect of the transaction. But even if I were able to weight all these aspects accurately, everyone else will have their own weight distribution. I don't care if this piece of art is late, but others probably would in their situation.
Daniel Kahneman's book "Noise" has some relevance here. The book distinguishes bias (consistently weighing your opinion in an irrelevant direction) vs. noise (jumping to a conclusion based on insufficient information). Reviewing someone is not only rife with biases, but it's also very noisy. People aren't motivated to write reviews consistently. (By the way, this is one reason why Amazon doesn't police reviews closely; leadership acknowledges that reviews are just one factor in a purchasing decision and shouldn't be over-analyzed.)
But as Adam points out in another comment, human recommendations of other humans is itself very noisy (and biased). Theoretically, I suppose that big decisions like hiring a contractor for home renovations should be done through friend-of-a-friend networks, because that human relationship prevents these sorts of problems, and it's one reason people cultivated those relationships in the past. But obviously we're not all in a position to do that for everything. It's a tough problem.
I feel like there are a lot of Web 2.0 companies that have monetized off of trust at scale like Airbnb, Uber, and any sharing economy company that became successful, even if they had a very strong incentive for making sure that every transaction was highly successful to the point of obfuscating failures.
I guess it’s hard at scale on the niche activities that don’t deal with food, shelter, or transportation. If I’m looking for the best coach, contractor, or doctor then it’s definitely hard to understand what’s right for me because there’s not enough data points across a wide range of human experience. Also because the complexity of the problem is much harder leading to variation in outcomes.
Great post, made me smile and nod throughout :)
I am in the process of building a house in Bangalore; I am going through sermons of how I should just "trust" the architect/builder with everything, despite no actions done to create trust. :)
Though I gave a detailed set of "requirements doc", there is no artifact for me to know they are on track. I know some are being modified -- maybe for good reasons, but without any discussions with me and my explicit consent. When I create a Google sheet to track those, there is heavy roll-of-the-eyes.
When it comes to "products", this is why I tend to trust the "Big Cos", as evil as they might be. Wish there were "services" too by Big Cos for individuals. (There are of course who just do the match making without any in house expertise -- they are dime a dozen and worse, I hear)
I've seen my parents, specially my father, deal with this issue in literally every aspect of his professional life, where he would have to resort to micromanaging most people from whom he would hire services . As a med student, I also find many examples of this "my way or the highway" mentality. Loved the piece! Am also a fan of George's writtings on Epistemink!
Interesting to think about social proof’s relationship with trust online - and how it’s gamed via online reviews, endorsements from influencers, and fake Instagram followers
Crowdsourcing is still the best option. Don't look at the good reviews, as they might have been gamed, but the bad.
Thank you for the post. I think your problem very nicely shows the value of “trust” or “low risk” and that the current platform systems don’t account for it. Coming from a more conservative background (my parents would definitely call around as well!), I kind of get it: The introduction of platforms focused on certain aspects of the services (price, availability) and ignored other that are less legible (good project management, less risk, “quality”). The natural equilibrium is that you get a lot of bad apples, as nobody pays for the quality control. (Reviews are voluntary, easily game-able and classic control instances like local newspapers and gossiping while waiting in line are often dead.)
So yeah, I guess paying up might be necessary (e.g. by trying to find the “honest builder” who - maybe out of personal values, principle, … is not on the platform), or to hope that the equilibrium somehow shifts, but that too will manifest in prices.
(One mechanism that goes into this way is - for products - YouTube reviews, where brands pay directly or indirectly for honest exposure. As long as social media has no way to localize, this is difficult for services, though - apart from franchises, that is, which are the other way to differentiate through price/“brand” again…)
On an off note, I feel this shows where the widespread current “experts”/“systems” framing is off, as so much of this boils again down to specific humans and their interactions, “community”, but not in an intellectual or esoteric sense, but the one you hint at. Paying your school mate because you know him, know he’ll give you the job as well and the gossip after he fucked up will hurt his business.