Smelling “ethics smells” when it looks ethical, but something smells off… Dorothea Salo

Programmers and “code smells” • Sometimes a given chunk of programming code works, but appears highly likely to cause a problem down the road. • Missing edge cases • Lots of cut-and-pasted code (unmaintainable after a certain point) • Cuts corners on security • Etc etc etc—ask a dozen programmers, get fi fty problems cited • These hints at poor/unstable/unmaintainable/hackable code are called “code smells.” • I ruthlessly appropriated this idea into “ethics smells.”

What’s an ethics smell? • A trope, rhetorical fl ourish, or line of reasoning that may look reasonable at fi rst glance, but hints at ethical fl aws. • Not everyone guilty of an ethics smell is a Bad Person doing Bad Things. It may not be that simple. • Sometimes ethics smells become part of prevailing discourse! Such that omitting or questioning them feels weird! • Sometimes people just haven’t done the work yet, or are seduced by whatever the latest shiny thing is. • Sometimes the ethical issues are legitimately hard to understand! • But yes, sometimes people do use these tropes etc. to forestall or de fl ect legitimate ethical critique. That’s not okay. It is, in fact, unethical.

Empty platitudes • E.g. “Your privacy and security are very important to us.” • Who even believes this now? We’re all like “prove it,” for good reason. • How can you tell a pleasant-seeming ethics platitude is empty? • Is any information given about actual actions in support of the ethics? If not, be suspicious. • If what they’re saying about the ethics issue confuses you… that may be intentional. Be suspicious… especially where they’re using way more words, or way more jargon, than seems warranted. • When in doubt, check into the person or organization’s track record. Is there reason to believe they will act (un)ethically? Are they self-serving? • Equitywashing is another common example. • “Of course my thing will be fair to and accessible by everybody!”

One dumpster fi re of bewildering ba ff l egab later… Higher-ed’s Surveillance Central!

“We’re good people! It’ll be fi ne! It’s fi ne because it’s us!” • Good is not a thing you can be. It’s only a thing you can DO. • When you stop doing good—especially when you start doing bad—you fall right off the high road. • This one is psychologically understandable; we all desperately want to believe we’re good people. • Ironically, insisting that we are good people no matter what we DO raises the chances we stop governing ourselves aright. • LIBRARIANS, WE ARE PRONE TO THIS ONE. • “To serve our patrons!” can become the door to evil. I’ve seen it.

No content

No content

Bluntly: no author on that paper should ever be allowed to work with data from living human beings again, especially not people in crisis! How. DARE. They.

Exact words • (see TVTropes.org on this, also “Insistent Terminology”) • Often used to justify ethically dubious behavior because fi fty pages of legalese supposedly spelled it out, and/or the victims “consented” to it. • “if you read the Exact Words of our novel-length privacy policy…” • Sometimes a distraction tactic; see e.g. Facebook: “the Cambridge Analytica thing wasn’t actually a breach!” • Technically true—Cambridge Analytica was allowed access to the data they abused by Facebook itself; nobody’s security was broken— but so what? • How is “Facebook allowed CA to use Facebook data to try to throw elections!” supposed to be an ethical improvement here?

No content

“But my thing is neutral!” • Librarians: we are SO GUILTY of this one. So very, very guilty. • It’s also rife in computer science. I don’t know why (and I doubt it’s a legacy from LIS), but I know from experience and reading that it’s so. • Often the subtext here is “how dare you judge me?” or “I shouldn’t have to think about the ethics of what I do!” • Often an attempt to evade ethical questions around power and oppression. • If it’s “neutral,” it can’t be racist or sexist or ableist or ageist or homophobic or transphobic OR GENOCIDAL (Facebook!!!), can it?! • Decades of STS, LIS, ethics work giving the lie to “neutrality.” I don’t have time to recount it all (ask me if you want readings).

The “neutral platform” Story from The Atlantic.

Own the good, ignore the bad • aka “because it helps [some] people, it must be in the clear ethically!” • This can be nakedly self-serving, of course. • And it fails consequentialist ethics forever! • Can also be power and privilege: “this thing helps some people (often ‘people just like me’), so all the people it hurts don’t matter!” • Sometimes it’s naïve optimism at work. • You may be able to tell the difference based on how long the responsible entity has been up to whatever the thing is. Are they still under the spell of the shiny? Or should they really know better by now? • I say this because you’ll want to tailor your counters. • Take the self-serving and privilege-leveraging down. No mercy. • Ask the optimists questions about how they’re handling their thing’s ethical issues. They don’t have to be gentle questions, necessarily!

This guy should know better. Remedial history/sociology-of-medicine class time! Story: Misha Angrist, “Do You Belong To You?” 2 January 2018. http://genomemag.com/do-you-belong-to-you/ Fair use asserted.

But what do we do with this?

Variants on “ignore the bad” • FOMOngering • “Everybody’s doing this!” (I have yet to see this be true when said aloud. It is invariably bullying or rationalization, not fact.) • “If you’re not doing this, you’re WrongBadBehind and you’ll die! Unethical-Thing is your only hope!” This is bullying. • Risks, what risks? Ethics, what ethics? • De fi nitely look for a self-serving (or at least self-justifying) motive here. • “We have an obligation to do this! How can we not?” • Often accompanied by a heartstring-tugging case study • Conspicuous by its absence: any discussion of risks or harms, except possibly for an Empty Platitude or two • (This is so common in learning analytics. So very, very common.)

No content

Variant: Scaremongering • “If you don’t do this dubiously-ethical thing, All The Bad Stuff will happen and it’ll be YOUR FAULT!” • So nakedly self-serving that you’d think people would see through it instantly… but if the Bad Stuff is bad enough and people are desperate enough to avoid it, we land here. • Sometimes the desire to Do Something, or Be Seen To Be Doing Something, overcomes common sense. • (Comes up frequently to justify surveillance because of crime-avoidance/safety.) • Also common in innovation-race discourse, e.g. AI. • “You can’t regulate our thing! We’ll lose the race to [enemy country]!”

No content

No content

“We can’t tell them what we do! Or let them say no to it! They just wouldn’t understand!” • This is invariably code for “we’re doing a Bad Thing and we don’t want people to know about it because they would object to it.” • I don’t love notice-and-consent. I will never love notice-and- consent. (See “Exact Words” slide for why.) • But keeping dodgy practices secret is even worse.

No content

“Those pesky ethics, always getting in the way of progress” • I don’t think I’ve ever seen this be so much as well- intended, never mind ethics-minded. It exists to delegitimize ethical concerns and those who hold them. • Seeing it? Question “progress,” loudly and often. • “Progress” is not always the word that will be used, of course; “innovation” is another common culprit. • Medicine: Primum non nocere = fi rst, do no harm. • Not the same as a con fl icting-ethics dilemma—that’s normal, to be expected, and absolutely a legitimate thing to think through.

When people tell you who they are, believe them! Oakleaf, Megan. 2018. “Library integration in institutional learning analytics.” https://library.educause.edu/-/media/ fi les/library/2018/11/liila.pdf

Variant: “Ethics are out of scope.” • Technology standardistas and startups are fond of this one. • I got into such an argument on Twitter once about whether OAIS should require digital archives to pay attention to risks to individuals or groups represented in a dataset, equivalently to “designated community” of data end-users… • The fi ght ended with an OAIS booster declaring repeatedly that this question was out of scope for OAIS. If so, then OAIS needs to fi x its scope, in my not-entirely-humble but trying-to-be-ethical opinion. • The IETF fi nally, fi nally, FINALLY repudiated this in 2019: https://datatracker.ietf.org/doc/draft-iab-for-the-users/ • “The IAB encourages the IETF to explicitly consider user impacts and points of view in any IETF work.” Whoa. That’s a change for the better.

“This problematic thing is a fact of life now.” • Zuckerberg is full of these. “Only connect.” • Ethics rule of thumb: If you start to sound like The Zuck, stop. • So many Big Data news stories and explainers start with this! • So much that I think it’s now one of those part-of-the-discourse things… • Sometimes, as with The Zuck, this is a naked attempt to set the terms of discourse to evade ethical responsibility. • In my experience, though, this one is especially likely to occur in the writing and work of people who are basically decent. • If that’s you, my advice is, do not cede ground without a fi ght. • Is the problematic thing really a fact of life? Is it truly unsubstitutable, unstoppable? Or is it just convenient for some people if we all think so?

Big assumption!

“Nobody told me I shouldn’t!” • the fl ip side of “I was ordered to” perhaps? Since Nuremberg at least, we’ve known that one’s out of bounds! • see also TVTropes “Ain’t No Rule” (often “Ain’t No Law/Regulation”) • So ethically lazy I can’t even. OWN YOUR ACTIONS. But I’ve seen it. • Assumes law/ethics systems are perfect and omniscient, such that if someone were really doing something bad, they’d be stopped. • This is nonsense, of course. Law lags bad behavior even more than ethics! • Not the same as “the ethics system here has gaps/loopholes; here’s how I navigated them, though I may have gotten it wrong.” • Where you see this, it is typically a thoughtful analysis that understands that where ethical guidance is not clear or there is no applicable guidance, we have a duty to think the ethics through on our own.

About that “gayface” study… From Metcalf, Jacob. 2017. “‘The study has been approved by the IRB’: Gayface AI, research hype and the pervasive data ethics gap.” https://medium.com/pervade-team/the-study-has-been-approved-by-the-irb-gayface-ai- research-hype-and-the-pervasive-data-ethics-ed76171b882c

“Who could have known this would be bad?” • Often a sign of someone who didn’t do their historical or ethical homework. Or said “talk to the hand!” to anyone who DID know. • Also common where the responsible entity is not inclusive. • See e.g. Google, which screwed this up repeatedly with search, image auto-tagging, Buzz, and Google+ (look up the “nymwars”). • Sometimes a sign of a new, naïvely optimistic person or fi eld. • Information security labors under a lot of technology standards and infrastructures that were poorly designed because nobody thought anybody would ever hack or misuse them. (See e.g.: email, BGP.) • I can grudgingly forgive this… for a little while. (For most technosocial phenomena… it’s way, way too late for any more of this.)

Story from The Atlantic.

Variant: “Who could have known this wouldn’t work?” • self-archiving in institutional repositories, MOOCs, AI in radiology (Geoffrey Hinton), cryptocurrency, NFTs, DAOs… • Usually someone knew it wasn’t going to work, and explained pretty clearly why not. • self-archiving: the major boosters were Cliff Lynch, Raym Crow, and Stevan Harnad; the one who knew was me, actually • (Salo 2009, “Innkeeper at the Roach Motel.” See also an angrier-and- wiser Salo 2013, “How to Scuttle a Scholarly Communication Initative.”) • Not coincidentally, the hypesters tend to be cis white men, and those who know it won’t work… aren’t. • Self-archiving did have one white female booster: Alma Swan. The rest were cis white men.

Variant: “It’s too early to tell if this thing I did will be bad. It might even be good!” • This variant often comes from academic researchers, particularly of the “tech is neutral!” variety. • Often code: “I don’t want to think about ethics! Don’t make me!” • If you see a lot of security/privacy experts and/or historians and/or non-cis-white-men head-desking over something… examine what they’re saying carefully. • The head-deskers will usually be vili fi ed — as Luddites, as anti- progress, as resentful, as Generally Bad People, whatever. • But my experience, for what it’s worth, is that these folks aren’t alarmists, just people who have seen this before and understand the patterns behind it.

No content

Variant: “Now I know! It’s really bad! Listen to me!” • Not an apology, usually — expect a lot of fauxpologies and “how could I have known?” nonsense — but a bid for continued (or even greater) attention. • These guys (and yeah, they’re practically always cis white guys) never credit the people (often not cis white guys) who had it right all along. • Indeed, they proceed to suck up all the air in the critique room, leaving no space for other voices. • Media outlets let them do this. That’s not okay either. Ask them why they had to learn this the hard way when there was plenty of evidence already! • Examples: Geoffrey Hinton (AI), Tristan Harris (everything ever, basically)

No content

About them, without them • (riff on “Nihil de nobis sine nobis,” which has been around a long time) • “We’re going to use this questionably-ethical thing to fi x all Their problems for Them!” • Whoever They are, They are either conspicuous by Their absence, or “represented” by one token person/case study viewed from a de fi cit model. • They are frequently in a less-powerful, less-privileged, less-voice-y position than the speaker. They may in fact have no option to refuse the intervention, which is questionably ethical all by itself. • So much wrong with this. So much. • Othering people is deeply uncool. So is disempowering people while ignoring their voices. • “De fi cit model”—They have problems (not strengths!), and We Know Better Than They Do how to fi x Them. • Often operating off stereotypes and super fi cialities—frequently blatantly incorrect ones—about Them.

No content

Variant: Problematic thing for thee, but not for me • Ethically-problematic thing is forced onto people with less social power, miraculously bypassing people with more. • Exceptionally common with surveillance and surveillance- based interventions • predictive analytics in government • surveillance in law enforcement • workplace surveillance (“bossware”) • learning analytics • the entire history of Facebook • Beware surveillance creep! Just because you’ve not yet been caught in a dragnet doesn’t mean you never will be.

No content

Variant: “I have nothing to hide! So you must not need privacy either!” • Missing the “from whom?” piece • Often conceptualized as “from law enforcement” or “from social media.” Privacy is much more complicated than that, or context collapse wouldn’t be so much of a problem. • Also a very narrow sense of context • My rejoinder is often “Oh? Cool. Hand over your wallet and unlocked phone, please, and the keys to your vehicle and dwelling.” • Also ignores bene fi ts of privacy: even when not strictly necessary, it is often useful and bene fi cial. • Also SO PRIVILEGED I CAN’T EVEN—er, falsely universalized • You won’t be persecuted for your gender identity, religion (or lack thereof), race/ethnicity/ancestry, sexual preferences, entertainment choices? Lucky you. Many of us are not in that place, and we matter too.

No content

I’m sure there are more! • These are just the ethics smells that either have been itching me forever, or that I happened to notice recently. • Contribute your own!

Thanks! This presentation copyright 2024 by Dorothea Salo. It is available under a Creative Commons Attribution 4.0 International license.