Surveilled #70–Another Facebook leak, AI bias and dark patterns
Issue 70
Another shorter issue this week, I didn’t come across anything that seemed to warrant the “one big story” treatment, although there are a couple of items in the “Six Links” that could well develop into one at some point.
Six Links
Some 500M records of Facebook users leaked, with email addresses and phone numbers–Among them, and just looking at Europe, the private info of EU Justice Commissioner Didier Reynders and Luxembourg Prime Minister Xavier Bettel. Not particularly surprising, but could be more dangerous than it seems with the inclusion of phone numbers (think two-factor authentication using SMS), and true to form, Facebook’s response to the incident leaves to be desired. The data was obtained through a vulnerability that was fixed in 2019, but individual users were not warned that their data might have been exposed. Read (Wired)
Southeast Asian platform app Grab is planning to go public–That is not surprising in itself, but the method is unusual: Grab plans to merge with a Special Purpose Acquisition Company (SPAC) launched by Altimeter Capital, a reputed late-stage venture capitalist. This is probably the highest profile deal involving a SPAC so far, just as the initial enthusiasm for SPACs seems to be waning. It would also be interesting to understand Grab’s motivation for dismissing a more conventional Initial Public Offering (IPO) or even a direct listing. Valuation might play a role, and the disastrous IPO of meal delivery app Deliveroo in London a couple of weeks ago probably didn’t help. Read (FT $)
The Apple vs Epic trial is unearthing all sorts of interesting insights–Epic is the company behind gaming sensation Fortnite, and objects to Apple’s App Store rules, claiming lock-in and discrimination. Among the millions of pages of evidence are some interesting insights into Apple’s strategy. App Store revenue and strategy for one, but also the positioning of iMessage as a device for lock-in to the iPhone. Despite some of the outrage that this seems to generate online, none of this is particularly surprising or egregious for a profit-making company... Read (The Verge)
Researchers are having difficulty measuring bias in AI algorithms–One of the reasons is that the datasets fed to the algorithms often omit explicit demographic characteristics, like date of birth/age, gender etc. The algorithm, however, may still proceed to show bias towards adjacent characteristics, for example, instead of “girls,” “people playing netball in college.” Afterwards, researchers can’t find this bias, because they don’t know the demographic characteristic to begin with. An example of a good intention that leads to unexpected complications. Read (ACM)
People are rushing into NFTs, but for some, the artworks are vanishing–Long story (involving the mechanics of the blockchain, Ethereum and NFTs) short, the complexity of collecting art this way prevents it from becoming a serious market, in my opinion. Read (Vice)
Good case study in “dark patterns”: the Trump presidential campaign–A dark pattern is a way to design a form, like a subscription or sign-up form, in such a way as to trick the user into buying or opting into things they probably don’t want. A classic example is the “receive email updates” option ticked by default, but there are far more sophisticated patterns out there, and the Trump campaign provides some great examples. Caveat emptor! Read