On Loyalty
Meet Jiro.
Jiro is perhaps the greatest sushi chef in Japan. He runs a tiny sushi restaurant in a Tokyo subway, which serves six people for dinner. Jiro and his team spend all day preparing sushi for dinner, and each morsel is the product of laborious planning, consummate skill and exquisite ingredients. The climax of the meal is a single piece of sashimi, and Jiro works incredibly hard to make sure that everything about it is perfect. In fact, each piece of sashimi costs Jiro a thousand dollars to make!
But Jiro has a problem - his guests love his sushi, but they do have preferences - some prefer salmon, and some prefer tuna. So when Jiro is planning his sashimi, should he serve salmon or tuna - or both? If Jiro wants to make his guests as happy as possible, he needs to serve them their favourite kind of fish - but how is he meant to know the preferences of his guests?
Jiro has two options. He could make six portions of tuna sashimi and six portions of salmon, and let his guests choose at dinner (of course, it would be awful if there weren’t enough of each kind of fish for everyone who wanted it); or when his guests made their booking, he could ask them whether they preferred salmon or tuna. Then he could go down to the Tsukiji fish market in the morning, and buy exactly the right amount of each kind of fish.
In the first scenario, Jiro has to spend twelve thousand dollars on preparing twelve pieces of sushi; but if he knows what his guests will want in advance, he can provide exactly the same product at half the price. Of course, his guests may find it slightly inconvenient to be asked in advance whether they prefer salmon or tuna; and they might even be willing to pay an extra thousand dollars a head to avoid that inconvenience. But in this framing, it's clear that their decision not to give Jiro information about their preferences increased costs and thus prices; it feels unreasonable.
Informational exchanges like the one I've just described are crucial to modern economies. If we can make it as efficient as possible for consumers to communicate their preferences to businesses, then the same goods can be provided at lower prices. This is exactly how just-in-time manufacturing techniques work. Originally developed in Japan, JIT means that components are delivered down the manufacturing chain exactly when they are needed, reducing storage requirements, depreciation, and time-to-money for everyone involved. This is in some ways a problem of practical logistics, but it’s really just about information - making sure that everyone in a process knows what is needed where and when. If an economy communicates information effectively between various players, then consumers get their goods more cheaply.
Here’s an argument about privacy and Tesco clubcards.
"In the current cost of living crisis, each day I read of or listen to media reporting of the increasing numbers of people who can't afford to put food on the table, who are looking for cheaper brands of basic foods, and what foods they can cut out. Kids going hungry.
At the same time, over the past few months I've noticed how supermarkets like Tesco's have increased their promotion of significant discounts on key foods when bought using the company's loyalty club card (now also a mobile app).
It's really got me thinking about the 'poverty of privacy'. For people struggling to put food on the table .. the loyalty card is quite appealing, especially when the Tesco 'Clubcard' is now also a mobile app offering convenience. But at what cost to people's privacy?
Great question. Pat Walshe, the author of these tweets, is currently Data Privacy Officer for Brave Browser and has worked in the field since 1998 - in contrast, I only started reading about algorithmic ethics in 2018. Furthermore, my current job involves serving mobile ads as profitably as possible, a profession significantly impacted by the privacy provisions in Apple's ATT policies. We should probably both be mindful the fact that, as Upton Sinclair once said, "It is difficult to get a man to understand something, when his salary depends on his not understanding it."
Despite our respective backgrounds, the issue at stake here is really just a trade like the one between Jiro and his customers. Consumers can trade information about their preferences for lower prices. Is it a good trade for these consumers?
On the one hand, if Tesco has more information about consumer shopping habits, their operations get more efficient. They will be able to optimise inventory, deliveries, pricing, placement, and so on and so forth. In theory, this reduces costs while providing better products to the consumer. In aggregate, Tesco has to hand down these benefits to the consumer because in most parts of the country, they operate in a highly competitive, highly price-sensitive, and fairly commoditised marketplace. Tesco Clubcards should mean lower prices and better products for consumers - a hard point to overlook given the current cost of living crisis.
Walshe doesn't actually acknowledge any of that though. It seems to me that he's part of a left discourse that assumes that almost all relationships between businesses and consumers are predatory, and that every time a consumer makes a purchase, they run the risk of being defrauded and exploited. Whenever the consumer isn’t being exploited, someone else in the supply chain is - this leads people to believe that there’s “no ethical consumption under capitalism”.
There are two ways that this trade could turn out to be bad for consumers. First, it's possible that actually, consumers don't see lower prices, because supermarkets siphon off profits for their shareholders. This is the criticism raised against energy companies when they report profits, but it seems incredibly unlikely a. in a sector as competitive as supermarkets and b. given that the entire premise of the trade is that consumers get lower prices. Walshe instead focuses on a second issue - the idea that there is a pernicious impact on consumers from sharing their data with supermarkets. Walshe marshals two points in support of this argument.
First, he goes through Tesco's Clubcard app on his (Android) phone, listing the SDKs found in the app. These SDKs don't actually reveal anything too surprising - in fact, I'm surprised the list isn't longer. AdMob, part of Google, have their SDK in most Android apps, while Adjust is just a mobile measurement partner, helping advertisers avoid fraud. Certainly it's not possible to argue based on the SDKs alone that the Tesco clubcard app presents a significant risk of harm to ordinary consumers. Walshe caricatures these SDKs as part of the #DataVampire - but could just as well be characterised as part of an online ecosystem that efficiently exchanges information to produce better and cheaper products for consumers.
The second, more interesting point raised by Walshe is the 2017 law review article Privacy, Poverty, and Big Data: A Matrix of Vulnerabilities for Poor Americans. The article explains the mechanism by which failures of online privacy cause harm to already-marginalised communities, in three distinct ways: algorithmic determined employment screening, higher education admissions, and predictive policing. Each of these case studies (especially predictive policing) have been well-studied, and the mechanism for harm is clear. The authors cite existing tools produced by companies like HireVue, and explain how networked privacy means that individuals may be held accountable for group-level statistics: if people from a certain social group are more likely to a. have lax privacy settings on their social media accounts and b. post unprofessional or offensive content on those accounts, then individuals from that social group may be held liable for other people's posts by an algorithm, and labeled as unreliable or untrustworthy - even if they themselves have done nothing wrong.It's important to note, however, that the algorithms in question are trained on social media, and that these systems already exist. But Walshe is using this paper to criticise the collection of an entirely different, significantly more innocuous and less personal form of data, without evidence that it is actually being used to cause harm to individuals.
Walshe himself notes that Tesco clubcards were first developed in 1994 - and yet in that time nobody has uncovered significant consumer harms arising from their use, despite a few fairly-high profile academic papers (this is probably the best law review article I’ve ever read) around predictive policing and credit-scoring. In fact, Martin Lewis, the well-respected cost-of-living crusader, declares on his website that "Tesco's Clubcard is one of the most generous loyalty schemes around, if used the right way." To me, supermarket loyalty cards are like a great success story of consumer capitalism, creating economic surplus in the form of lower prices by efficiently exchanging information between customers and supermarkets. While there are important reasons to safeguard privacy in certain contexts, this seems like a good trade for consumers.