7

Many of the posts regarding AI mention the same issue regarding AI Bots: "Online, I can't tell the difference between a human and machine response.

I don't see a solution to this problem that doesn't seriously curtail anonymity on the Internet. A reasonable solution seems to be the forced registration of AI software or the elimination of anonymous Users. Are there philosophical arguments regarding the right of humans to be anonymous?

Edit: By the responses, it's seems the pros of anonymity vastly outweigh the concerns raised by the inability to distinguish between a human and a Bot online. Until there is evidence that this activity is any more detrimental than the standard online human deception, there is nothing broken to fix.

Thanks to all for the thoughtful responses and comments.

3
  • Comments have been moved to chat; please do not continue the discussion here. Before posting a comment below this one, please review the purposes of comments. Comments that do not request clarification or suggest improvements usually belong as an answer, on Philosophy Meta, or in Philosophy Chat. Comments continuing discussion may be removed.
    – Philip Klöcking
    Aug 3, 2023 at 12:28
  • 1
    Is this philosophy? It feels like primarily a question of politics or policy, and the answers reflect that.
    – Kaia
    Aug 3, 2023 at 16:20
  • @Kaia I truly don't know. It's a muddled mess of psychology, philosophy, and misinformation. I've gotten mixed reviews but I've received 9 excellent answers, 5000 views, and many fine comments. I've hit a nerve and I'm still curious where it leads.
    – user64314
    Aug 3, 2023 at 16:42

8 Answers 8

11

Yes, but not in the way your question implies it. Automized language models will provide a huge challenge for anonymity, but likely not because we can't tell them apart, but rather because they are probably able to tell us apart. If you write anything more complex than "yes" or "no" you'll exhibit some sort of style that is unique to you or at least unique within a given context and that persists between different "identities" (online accounts).

So if only 1 of your accounts is linked to your identity or if you have enough linked accounts or data within an account, one can probably harvest enough information about you to make a profile about who you are. And either that already links to a person or is enough information to scan various databases and networks and find the person that fits. If you have enough parameters everyone is unique.

On the other hand it might become increasingly hard to even verify that you are a person. I mean how would you do that? Upload a picture? Deep fake! Upload a blood sample? Just search google scholar for how these look like and fake one! Social Security numbers? Would you want that to be public? You could make a registry that links names and tokens and confirms them. And a bot net could take an common name and brute force a bunch of numbers till it fits.

Everything needs to be digitized in order for easy sharing and accessing and everything that is digitized can be forged.

6
  • 4
    "How would you do that? ...Social Security numbers?" If you trust the entity that issues Social Security numbers, but don't want to make something public that could be used to forge your identity, cryptography has got your back: let the SS-assigning entity also assign you a digital signature key. For an approach that requires less trust and less government cooperation (but perhaps gives less guarantees), check out Keybase, which is at its core about connecting "you" with an unforgeable identity. Aug 1, 2023 at 14:23
  • 1
    @DanielWagner I would suggest Keyoxide makes your point better.
    – wizzwizz4
    Aug 1, 2023 at 15:31
  • Nice answer. Apart from psychology (it's creepy), is there any real evidence to suggest that the inability to distinguish between a human and a Bot online is any more detrimental than being deceived by a human using a fake persona? That kind of criminal activity hasn't curtailed anonymity.
    – user64314
    Aug 2, 2023 at 1:41
  • 1
    @StevanV.Saban - I think there's already some evidence that AI generated spam has a far higher success rate than non Aug 2, 2023 at 10:30
  • 1
    @StevanV.Saban Not sure it is more detrimental than being deceived by a fake persona, but as ScottishTapWater has said, it might be easier to deceive you as it has more data about how people speak and more data about you in particular. A human attacker would either target one particular victim with huge research effort or a large number of people with a broad unpersonal statement, an AI might do both large scale and personalized.
    – haxor789
    Aug 2, 2023 at 11:18
6

I think it was always quite a naive assumption that internet posting is a fair reflection of a public voice.

Computing is a technology, and the online voice has always reflected the perspectives of those in possession of that technical capability, whether the original ARPAnet or the global middle class utopia of Tumblr.

What we’re seeing now is a different take on this, in that governance of that voice is shaped much more directly by the construction and use of large language models, but the principle is very much the same:

Everyone is a dog on the internet.

2
  • Let the user beware.
    – user64314
    Jul 31, 2023 at 18:42
  • 2
    @StevanV.Saban, very much so, but there is also something to be said for a kind of User Group mentality. If the technology is more of a network than a product then we’re getting somewhere - the problems start to come in where the commodification and streamlining of it leads to monopoly control.
    – Paul Ross
    Jul 31, 2023 at 19:04
6

The question we're facing isn't just about distinguishing between silicon and flesh, or between the creations of our machines and our own creations. It's a question about the nature of identity, about the right to speak and to be heard, and about the balance between transparency and anonymity.

There are those who argue for forced registration of AI or the elimination of anonymous users. On the surface, these might seem like reasonable solutions. But we must be wary of any solution that requires us to cede more control to authorities, or that infringes on our rights to speak freely and anonymously.

Anonymity, after all, can be a shield for the persecuted, a cloak for whistleblowers, a necessary disguise for those living under repressive regimes. It's easy to forget, amidst our comfortable debates about AI and identity, that for some people, anonymity is a matter of life and death.

1
  • 3
    I agree. Anonymity is very powerful and excellent for all the reasons you mention. But there is a dark side. Anonymity can be used to act without consequences. Chatbots and Anonymity may be the most powerful tool for propaganda yet. An equilibrium can be established between the pros and cons but it will come at a cost.
    – user64314
    Aug 1, 2023 at 15:42
5

Privacy activist Edward Snowden once said:

Arguing that you don't care about the right to privacy because you have nothing to hide is no different than saying you don't care about free speech because you have nothing to say.

Perhaps you don't need anonymity. Perhaps you are OK with everyone being able to attribute everything you write on the Internet to you. But not everyone might be OK with that. For example:

  • They live under an oppressive regime that punishes them for speaking out against it.
  • They are afraid they might one day live under an oppressive regime that will get rid of people who showed disagreement with their agenda in the past.
  • They live in a society where certain views are shunned and lead to social ostracism.
  • They want to publish information that is important for people to know, but which other people might want to remain secret. And those people might be able to retaliate against the speaker.
  • They want to share experiences about things that are embarrassing to them.

Without anonymity, those people would not be able to do any of these things. Which is not just bad for them, but also bad for you, because you have less useful information and less diverse viewpoints available which allow you to make sense of the world around you.

8
  • 2
    These scenarios all focus on good actors seeking to avoid unjustified retribution. Should this perhaps be weighed against the significantly more common inverse scenario of bad actors seeking to avoid justified retribution?
    – eclipz905
    Aug 2, 2023 at 17:44
  • While anonymity may be a right for people in order to ensure free speech, it is not a right for an autonomous machine and now you can't tell the difference between a human and an autonomous machine online. It is the ability for a machine to make autonomous decisions that forces different thinking about online anonymity.
    – user64314
    Aug 3, 2023 at 1:39
  • 1
    @eclipz905 actors are good or bad depending on their influence on the ruling structures, not on themselves.
    – EarlGrey
    Aug 3, 2023 at 9:49
  • 1
    @Philipp. If anonymity is to protect free speech, then isn't giving anonymity to a machine essentially giving the right of free speech to a machine?
    – user64314
    Aug 3, 2023 at 18:43
  • 1
    @eclipz905 who determines who is a 'bad actor' and what is 'justified retribution', if indeed such a thing even exists?
    – user
    Aug 3, 2023 at 22:15
3

Are there philosophical arguments regarding the right of humans to be anonymous?

Certainly, but like many things, they raise just as many questions as they answer. I'll focus on the ethics for or against a right of anonymity.


Under Kantian ethics, we'd look at universibility, and ask if it's logically possible to live in any of the following worlds:

  • A world where no one has privacy. Everyone knows everything about everyone else. What they did, what they want, what they are planning to do.

  • A world where everyone has perfect privacy. People are unidentifiable and completely interchangeable.

  • A world where everyone has some parts of their lives that are public and other parts that are private. Some of their actions are known publicly, and others are anonymous.

After some thought, I think none of those lead to logical contradictions. So, we wouldn't have a perfect duty to follow one course over another.

But, do we have an Imperfect Duty? Are any of those logical worlds, a world that we would not like to live in? A subjective preference of humankind?


Under Contracturalism, you might have a right to anonymity because we've agreed to it, to make society a good place for us all. But, if the situation changes, you might just need to change the contract. So, if we all agree to give up anonymity, then that's the right thing.


With Utilitarianism or Consequentialism, you'd have to ask whether the world is actually a better place with human anonymity? or without it? Is there more suffering? (more bias against minorities? identity theft? lying?) or more pleasure? (truth, trust).


Or under virtue ethics, you'd have to ask whether doing things anonymously (occasionally? or all the time?) is a good character trait. Is it good to be a person that gives to the poor anonymously? or is it okay to give to the poor even if everyone knows who you are and what you are doing?

2

This is less about the philosophical aspects and more about getting the practical aspects right, because it's not immediately clear to me that the possibilities you sketch are the only ones.

First of all, in some cases it's not a problem if a user might be a bot, for example if all that that user is doing is reading others' content (assuming there's no problem with that content being scraped by a bot). So in this case we can still have anonymity. But I think what you have in mind is settings where this isn't the case, for example because the user is creating content such as text and we don't want bot-created text.

Now it may still be possible to have effective CAPTCHAs (perhaps relying on hard vision problems) to verify that users are humans. Perhaps those humans would post text that is generated by ChatGPT, but at that point humans are choosing to submit that specific content, so that may not be a problem.

Perhaps at some point AI is good enough that there are no longer any effective CAPTCHAs. At that point, it would not be possible to (automatically) tell humans from bots based on their behavior. In principle, one might yet have constructions such as "trusted humanity verifiers" -- i.e., trusted organizations that check that someone is a human being (perhaps the human needs to show up in person at the organization) without recording identifying information about that human, and then through the trusted organization the human is allowed to post content while still having anonymity.

1
  • 1
    Do you consider ChatGPT a reference to information or a source of information? This is where it gets blurry for me.
    – user64314
    Aug 1, 2023 at 15:09
2

I do not see any advanced AI tools as a particularly serious threat to anonymity in online communications. We are well past that phase in the war on misinformation.

You seem to be concerned by the use of chatbots as tool for [political] propaganda, but it is Facebook, not 4chan that is the primary target for such large-scale campaigns. It is relatively trivial for state-level actors to fabricate online identities. Thus, placing citizens under strict online control would be truly Orwellian, all while troll farms would still thrive (and there's nothing to prevent a given government from recruiting actual, verifiable humans to post online!).

Trust is a core feature of human social interaction. Yet, it is easily exploitable, and one does not need advanced AI for that: stories of con men are countless, and modern media has mastered the art of clickbait. What differentiates AI from human writers is productivity and feasibility to mass-target a people exploiting their specific fears and desires... But I would argue that if someone habitually takes the bait, sooner or later they would fall prey to much cruder approaches anyway.

I am Russian, and I have friends, relatives and colleagues exposed to both sides of political propaganda with mixed results. The main differentiator between those who easily fall to things akin to QAnon conspiracies and those who don't is critical thinking. And a bit of digital hygiene, too, I suppose.

In this regard, the rapid progress in LLMs is merely a part of a broader issue that is ethics of social media and algorithmic recommendations. If the key metric is user engagement with the platform, how much is left of free will? Do people exercise it when following Siri's recommendations of what to have for dinner, what to wear, and how to spend their weekends? The humanity has fought hard to bear the burden of responsibility for one is choices and the freedom to make them; now it is given away without much of a fight.

Going on a bit of a tangent here, but one other major thing we have lost since the Cold War, albeit not entirely, is the trust in public institutions and science/engineering as a whole. It would be much, much easier to fight misinformation online if people would refer to several authoritative, institutional sources instead of going with what their online friend pete93 has to tell them. After all, pete93 had a stunt 2 years ago working as janitor in NASA, they could not possibly be wrong!

Jokes aside, the main ethical question regarding misinformation online has to do with responsibility, and in my humble view, the part of the burden the general public can't bear should be squarely placed on public institutions with maximal transparency. More importantly, said general public has to accept more responsibility for being deceived than it does right now: it may be a malicious actor who sought to exploit you, and you might rightfully be wishing to live in a safe world where it does not happen... But this is simply not the case right now, and it hardly ever was throughout the history.

In short,

Things that help creating and spreading misinformation online:

  • AI
  • Conspiracy theorists
  • Lack of education

Things that help fighting misinformation:

  • Critical thinking
  • Transparency
  • Trust in public institutions

Things that do not help fighting misinformation:

  • Biometric IDs online
  • Thought policing

As @haxor798 has hinted at, linking every single account makes targeting people online actually easier, which, I would argue, would only serve propaganda makers, harassers and trolls. There is little to gain and a lot to lose here.

In contrast, providing better informational infrastructure, educating the public on matters of misinformation online and digital hygiene might alleviate the issue. After Montessori, I must repeat - "Establishing lasting peace is the work of education". And so is establishing safety.

3
  • Trust in public institutions? Do you trust government press releases? Are you suggesting citizens should?
    – Rushi
    Aug 3, 2023 at 7:13
  • @Rushi Public institutions are not limited to those with heavy political involvement - and yes, I think one should put more faith in them over what their crazy uncle said. Consider Sharpiegate - there are professional meteorologists doing their jobs and, ideally, you want to be able to just rely on their competence. If public institutions start to be used to further a political agenda, it creates (rightful) mistrust, which is overall extremely dangerous. If you can't trust anyone, citizens end up trusting whatever "their" guys are saying - and there is a lot of that happening right now.
    – Lodinn
    Aug 8, 2023 at 3:16
  • That is, it is better when teams of experts may be relied upon - but this requires work and preventing tampering driven by e.g. political forces. And it it governments' job to not have citizens second-guessing them every single time, while independent/adversarial observers are free to criticize their statements. When said critique is available to the public and the public deems government's statements at least mostly reliable, the system is working.
    – Lodinn
    Aug 8, 2023 at 3:20
0

I'm more bothered (maybe not just perplexed) by humans pretending to be humans they are not, than robots pretending to be human, and yet see no reason to demand that humans prove they have e.g. read Spinoza, before answering a question on Spinoza.

There's a difference between manipulation and anonymity in general (these bots are representing someone). What matters - IMHO - is less that we know which interactions are human than openness about why some of our interactions are not human. I probably have read Popper on the open society, but it's not limited to right wing capitalists (there was an 'Open Marxism' even).

Karl Popper defined the open society as one "in which an individual is confronted with personal decisions"

An attitude to institutions whereby we rationally decide their consequences for ourselves and our persons: presumably - though this is the key point I'm making - it's impossible if we don't know what other parts of society do (closed societies are tribal or seamless biological like entities).

Anyway, how can we struggle with and judge other individuals, and their ideas, in a society in which we don't know how they act or their moral attitudes?

Investor and philanthropist George Soros, a self-described follower of Karl Popper" [claimed]... "Politicians will respect, rather than manipulate, reality only if the public cares about the truth and punishes politicians when it catches them in deliberate deception."

It doesn't take a genius to notice that large scale deceptions are more likely in a closed society, and presumably if we care about the truth we want to know why we are or have been tricked (wellness aside).

1
  • Nearly all the pros for anonymity involve the protection of free speech. If the sole purpose of anonymity is to protect free speech, then isn't granting anonymity to a machine essentially giving the right of free speech to a machine?
    – user64314
    Aug 3, 2023 at 19:40

You must log in to answer this question.