If I were to build a bot I wouldn’t even touch the API. Why not simulate everything through a headless browser and stay under the radar. Then build a farm of those simulations and you can publish/promote everything you desire pretty much undetectable.

> If I were to build a bot I wouldn’t even touch the API.

If you were to build a malicious bot, sure. If you were to build a bot to tweet about the weather in Sheffield, you might as well use the API.

> If you were to build a bot to tweet about the weather in Sheffield, you might as well use the API.

I don't think twitter is that place and it hasn't been for a while if it ever was.

I just watched this video again...

https://youtu.be/ddO9idmax0o

To me, the draw of twitter was you could text messages (SMS) to 40404 or get texts from 40404 from people I choose to follow. I had unlimited minutes and unlimited texts but I didn't have an app phone back in 2008 when I signed up for Twitter. The world has changed since then...

I'd say if you were to build a bot to tweet about the weather in Sheffield, you are probably better off setting up your own server to write blog posts or something like that? Maybe look into microdata or RSS or something too?

It's very common to build bots for Twitter. They post on Twitter because there's an audience there. I'm sure there's lots of places I could get an rss weather subscription, but I stopped using rss.

This is a Twitter account that started relatively recently - I like it. https://twitter.com/_restaurant_bot

This sort of bot has had a good place on Twitter for at least a decade, going back to @everyword.

Yeah but the bots that people complain about are the malicious bots that try to pass as humans.

Those become a lot easier to work with when you differentiate them from good-faith bots.

Discord does something similar, although I feel they've gone way too far with policing it.

The idea was about opening the API, and this suggestion was to argument that it makes easier for the malicious bots. So this was suggested to make the API less valuable for that purpose.

I think the only problem would be if this API was used by non-bots (perhaps some relays) and the tag loses its meaning.

I think we need to expand our vocabulary a little on this subject. A lot of people seem to think that these "bots" are just limited to "buy crypto today" type posters.

What they don't realise is that the comment they are reading right now could well have been written by a bot.

His scope covers everything you mentioned, you are the only one limiting it yourself.

I’m not on Twitter, but I can think a myriad of cases where I wouldn’t want to touch the API without my case being malicious. Like extracting data from the platform, or monitor specific accounts, or monitor trends, or whatever. It doesn’t have to necessarily be a troll farm, or a fake users service. I’m sure the API allows most, if not all of these, but there would be usage quota. You bypass it, and then you can even start a business around it. There are thousands of startups out there providing services for social media monitoring, and I seriously doubt all of them are going through the official API.

But not every tweet through the api would be bots right? Like someone using a custom twitter client as a single user shouldn't get that label, no?

That's going to be tricky too because I could imagine someone offering a service to actors and other public people to help manage their twitter accounts (maybe discourage them from posting stupid shit at 1 am. I'm looking at you, Elon), in which case you wouldn't even have a 1:1 ratio.

The usual solution, historically, it to provide different levels of service based on your account type. The couple of times I've worked somewhere that the admins had a disjointed set of servers from the users was really a huge stress relief. You want admins to 'see' what the user sees, but to be able to work even if the site is being DOS'ed. And ideally you want users to be able to use the site even if the admins manage to DOS themselves (admin functions tend to have a higher fanout than end user functions, so the danger of catastrophic fanout is always higher).

One option would be to ban custom Twitter clients. As a bonus, enforcing official apps opens more possibilities for monetization. Twitter could even release a special app for advertising agencies who manage multiple accounts on behalf of their customers and charge appropriate subscription.

As a bonus? Maybe for Twitter, but as a user, I would not care for this at all.

My take is banning custom clients should be deemed an ADA violation.

You should be allowed to create your own client for any service you want, that is customized to your input needs.

How are you going to tell?

They can distribute keys associated to an identity and have humans review the apps or bot purpose and content, but charge for the review.

is there anything stopping me from writing a browser extension to make posts on the site automatically? i did work for a guy on that sort of setup spamming used car dealership ads, and it may have been a hastle keeping it running but we got plenty of ads out.

nothing other than the fact that if they find you doing this, they'll likely ban your account.

If the private keys are distributed inside an approved client app then any competent hacker can decompile the binary and extract the keys. At that point the keys can be used by bots or other unauthorized apps, and there's no reliable way for the server to distinguish them.

> Like someone using a custom twitter client as a single user shouldn't get that label, no?

Why not? If it's a niche client not many people use such false positive would be rare. Also such people probably don't have huge audirnce and they and their audiences won't care.

And it might be deterrent against people trying to build popular alternative client that might compete with main twitter app.

I don't think aiming for 100% accurate label is on the table.

> And it might be deterrent against people trying to build popular alternative client that might compete with main twitter app.

That's horrible for users.

Twitter's iOS app is a third-party app they bought because it was so much better than their own at the time. https://en.wikipedia.org/wiki/Tweetie

That was its origin, yes; but that was 12 years ago. The current official client is unrecognizable from Tweetie.

Yeah, but that was then and this is now. Thinghs kinda settled down since then.

The real danger of opening the API and marking all tweets made through API as BOT is that this label looses impact if some alternative client that uses API raises to prominence.

I'd say that current setup where users have no indication that something might have been posted by bot is way more horrible.

Humans are really creative. One potential outcome is a few well known ppl start using a 3rd party client, tell their followers that’s why it says bot, and the badge loses its meaning.

This is just kicking the can down the next level on the issue. They can't tell which bots are bots so what difference does it make it if the solution isn't to ban them but to just limit them? The problem is identifying the bots.

Without an API, the default "I want to write a bot for Twitter" path forward is to automate the human UI -- to some extent, this makes you look like somebody who is trying to evade bot detection.

I suspect there are lots of people in the class of "wants to do something interesting, isn't particularly interested in being malicious but doesn't mind breaking the rules." Maybe an official bot API would dry up that ecosystem?

I tweet using the iPhone app Tweetbot. Would my tweets be marked as a bot?

Do you identify as a bot?

So you're saying "bot identification" would be an opt-in feature? That would go well.

I'm pretty sure he was making a right wing juvenile comment

With an open API, you even open up the door to "moderation bots" that autoban users from your timeline if they appear to be bots

Hmm...this may turn into "Ban speech from my account that's promoting right/left wing agenda" moderation bots

Let the bot wars begin

I'm not going to lie, "bot wars" sounds a lot more fun than current Twitter. I want to play.

Those already do kind of exist, I recall scripts existing which can be used to block users from lists, and inevitably those were used for reinforcing echo chambers.

On the other hand, providing users with tools to curate their timelines is much better than having the company shove some arbitrary rulings down everyone's throats.

I think the "bot" versus "real" account differentiation is quite irrelevant in practice.

For example, every now and then I check what some elected Canadian politicians have tweeted recently.

Even though these are Twitter accounts that would probably be considered "real", and they often have a check mark, pretty much everything they tweet is, in effect, following a script.

Often, these tweets are just a useless acknowledgement of some obscure holiday celebrated by some unrelated ethnic or religious group that they want the votes of, or they're tweets delivering pointless congratulations for an irrelevant "accomplishment", or they're tweets regurgitating some very specific party talking point that a bunch of other representatives from the same party are simultaneously regurgitating.

These accounts are seen as "real" accounts corresponding to "real" people, yet what they produce is consistently some of the most artificial content I've ever seen.

That's an unusual example. Most posts on twitter are not formulaic recognition of holidays by politicians.

In general, not knowing if a machine or a human wrote a twitter post is a bad thing.

There’s many variations of this.

News outlets and fans tweeting sports results. Public event celebration (new year, xmas, father day etc.). The vast majority of these tweets could be managed by bots and we wouldn’t know the difference.

I actually follow many bots on twitter knowingly, but they could be people in disguise I wouldn’t care.

In my opinion, the content of a tweet (is it spam ? is it abusing the platform ?) matters way more than wether it’s a bot or a human doing the spamming/abuse.

> In my opinion, the content of a tweet (is it spam ? is it abusing the platform ?) matters way more than wether it’s a bot or a human doing the spamming/abuse.

What about replies and retweets? EX: If you have high quality (or good faith) content that's supported/attacked by bots vs real people.

I'm posting this only to further decrease the upvote/comment ratio.

I kinda wish engagement had less effect on content visibility and having a swarm of bots "attacking" a comment wasn't decisive.

Talking about HN comments, some have extremely bad but feel good takes that will garner a huge first wave of upvotes, to then get beaten down as it got more attention, to perhaps get some more sympathy votes as timezones shift.

The interesting part to me is the comment can go all the cycles because it stays there, is still visible even when flagged to death and can be rescued.

That's also how I read reddit, with downvoted to death comments and posts in the timeline. People who care enough should probably do the same for Twitter as well.

Disclaimer: I am not politicaly engaged so have less exposure to the nastier battles where thousands of people shout a each other. I still don't understand why anyone would go there in their free time.

I think TikTok nailed this. They will promote posts to their home page that do not have lots of engagement but but are high quality for the site.

Why do you want to know if a machine or human wrote a tweet? What difference does it make, and where do you draw the line? If I use Grammarly to correct my language in a tweet then did a machine write it? How about if I generate text using GPT-3 and then manually edit it?

Please define your terms.

One is more labour intensive and more expensive. The other can be done at scale for cheap. The negative impacts most likely don’t scale linearly, so volume is an outsized concern.

[not knowing if a machine or a human wrote a twitter post is a bad thing]

And what difference does it make, wether some content was written by an algorithm or it was written by professional staff, when the objective might be the same: to misguide / influence you to act in someone else's favour

A brief conversation is usually all it takes me to notice even the more elaborate bots. I would share my methods, but I'd prefer they not figure it out too quickly. Suffice it to say bots just don't behave the way a human would.

Most politicians are boring. So what? Just don’t follow them.

The problem with bots is not that they are “artificial sounding”. They can and are used for scaming people, they can and are used for manipulating the public discourse.

The problems with bots is they jack up advertisers cost and don't buy things and cannot be influenced. Follow the money.

There are bots on every platform and advertisers are well aware of them.

They simply factor them in as a cost of doing business.

some will also increase engagement though

I think the problem is less that many official accounts are "boring" and more that they are run by staff and don't have the feel of direct communication that makes Twitter interesting to begin with.

Politician social media accounts are mostly run be teams of staffers who are literally paid to manipulate the public discourse. They are no different from bots.

We've arrived to a very satisfying conclusion here: politics has always (mostly) been the biggest of all scams. Wether its a team of script writing artists, or the state of the art machine generated content, its the same worthless junk.

You can grade the junk by halflife.

The purpose of a BOT tag isn’t to indicate quality or intent, it’s not even necessarily derogatory. Real people can be asinine and bots can be useful and informative. It’s to indicate the source was a bot. That’s still useful information.

A BOT tag is a great idea. They should also add the option for authors to disable BOT replies in their threads. Best of both worlds.

Excellent idea, with BOT replies disabled by default please. Mostly people don't want BOT posts.

Useful how? What am I supposed to do with that information? Is it actionable?

Like a politician?

Everything is politics. Without political skills humans wouldn't have landed on the Moon, but it took the shock of a headshot and the legacy of that memory which eventually faded to just about nothing in ten years.

Man, this “manipulating the public discourse” boogeyman has really taken root in people’s brains. The CIA mind control research paid off after all.

Is it not fact that the manipulation of public discourse using bots on social media has taken place at scale?[0][1][2]

When U.S. Presidential elections can be significantly swayed by information warfare campaigns run by malign actors, I can't help but feel that dismissals in such a context are rather aloof, if not out of touch.

[0] https://intelligence.house.gov/social-media-content/

[1] https://www.dni.gov/files/ODNI/documents/assessments/ICA-dec...

[2] https://www.propublica.org/article/infamous-russian-troll-fa...

There's a difference between tried to manipulate and manipulated. We don't know what effect they had. I suspect a miniscule one.

If their efforts yielded minimal results, somebody else must have had success. Online discord seems to have been sown somehow.

Online discord is being sown because there is real world discord. It’s not coming out of thin air. The college educated upper middle class is increasingly pulling away economically from the middle class, and is being increasingly assertive about their value system which is increasingly diverging from the traditional values held by most of the rest of the country. Institutions are breaking down in different parts of the country in different ways—both red America and blue America are less likely to attend church than before, but what they do instead is creating greater gaps between them. Immigration and globalization are bringing an increased pace of change, and the New York Times is calling anyone who objects to that change bigots and racists.

On what basis exactly?

I can't speak for JamesBarney's reasons, but agree with them that it's probably minimal.

Facebook and Twitter have better funded research groups focusing just on the sub-task of knowing what my current state is (a necessary part of being able to tell if you've changed my preferences), and for advertising purposes I am categorised as being into nouns I don't recognise, foods and sports which repulse me, and so on.

I expect the three letter agencies to mix me up with the horror film director I share a name with, and also to confuse me with three pet dogs and two pet cats.

Because there's been lots of research on how much advertising affects election outcomes and it's minimal.( At the federal/state level, in smaller races where voters lack information about either candidate it can be slightly more influential)

I can't imagine a Russian internet trolls with much a smaller reach and lower budget are going to be that much effective than best political ad minds of America at changing our opinions. (For Russian governmental organizational competency see Ukraine)

And then anecdotally I could have predicted every person's i knows vote a year before based on how they felt about woke culture and their party identification.

What’s your evidence for the notion that the presidential campaign was “swayed” by “information warfare?”

The reports that I linked from the U.S. House of Representatives Permanent Select Committee on Intelligence, and the National Intelligence Council.

What degree of sway is up for debate, but given the documented scale and scope of the efforts, it certainly wasn't zero.

Ah yes, the "Russian Troll" conspiracy theory. Gonna get a lot of big mad downvotes for this one because people actually believe this nonsense.

The public has been able to be swayed by media for a very long time. The CIA in particular has been deeply involved in it [0]. For some reason this was deemed okay, or at the very least acceptable, by the public. The only difference between the CIA controlled media cited and social media is the lack of US government control.

I do appreciate the irony here. That every other election was so-called "free and fair" yet an absolute wildcard gets in for 2016 and the deepest level of coping by every agency in the union was witnessed. I am not a fan of Trump, but if elections are truly "free and fair" then 2016 was legitimate. If 2016 wasn't legitimate, then why would anyone have any reason to believe 2020 was legitimate? I'm not intending to start a conspiracy theory fight here - but rather raise the question that is trying to be asked. If "foreign actors" elected a president in 2016, then whose to say they didn't in 2020? Or 2008? Or 2004? Who can you believe? I'd argue you couldn't believe any agency who might have an interest in having you believe otherwise. Maybe our elections are actually selections and in 2016 the powers that be lost the information war.

[0] https://en.wikipedia.org/wiki/CIA_influence_on_public_opinio...

Is it possible that both CIA intervention into public discourse and foreign entities with unknown interests with reach far greater than the largest media outlets in the country[1] are both not ideal? And that one does not excuse the other?

1: https://www.technologyreview.com/2021/09/16/1035851/facebook...

It’s possible to believe that the 2016 result as legitimate, and that Russian Psyops influenced the result. After all it’s incumbent on the electorate to infirm themselves and employ critical thinking in the information they consume, but at the end of the day it’s their choice. That doesn’t mean Russian Psyops is ok, it doesn’t mean CIA Psyops is ok (while noting that you didn’t actually say that they are).

As an external observer (Brit) I deplore both those who try to malign the legitimacy of the 2016 result, and the 2020 result. Even so, I hold a special contempt in my heart for those Americans trying to excuse, justify or even in some cases explicitly encourage foreign covert influence on US elections and obstruct efforts to counteract it. Those people are not even borderline traitors and enemies of democracy, they’re all the way over the line. All IMHO, and it’s not my country.

I remember back then that there was also evidence that China ran some kind of bot campaign[1] but no one really cared and it was never brought up. Ironically your second source says that but again, no one seems to talk about Chinese bot farms or election interference strangely. That same source talks about all the countries that do this. I assume this is pretty routine and the US probably does it in other countries. I don't know why Russia is focused on for this in 2016. I have my guesses.

But then there were investigations into the Russian election swaying campaign. I remember Facebook said that there was no evidence the ads they bought had any impact on anything[2]. There were also groups the trolls made. It's completely unclear what conclusion to draw from them.[3] The article says people mostly didn't follow the groups but that their posts reached people. What conclusion can be drawn from this? Is there any evidence the posts changed people's opinions? Also, one of the weird and counter intuitive things about it is they targeted left wing people.[4]

So I don't see any evidence that an election was significantly swayed. I didn't see it in those sources. I am not trying to make this a left/right thing. I am just saying that yes, there are bot and troll campaigns all over the place. But I don't know if we should conclude that they significantly swayed an election. It's just something to keep in mind. I brought up the China thing because I would think if it was actually a concern then people would have talked about it but how many knew that China, or Venezuela or Iran, were involved in this?

[1]: https://www.reuters.com/article/us-usa-election-china/trump-...

[2]: https://fortune.com/2017/11/01/facebook-google-twitter-russi...

[3]: https://www.technologyreview.com/2021/09/16/1035851/facebook...

[4]: https://www.reuters.com/article/usa-election-facebook-russia...

>So I don't see any evidence that an election was significantly swayed.

Well, in my defense I did say can be swayed. However, I do think there was serious damage done when you take the full scope of Russian influence operations into account.[0]

That said, considering just the troll farm influence alone in an election scenario with tight margins—the 2000 election, for example—it's rather easy to see how it could make the difference.

It's taken seriously for good reason.

[0] https://en.wikipedia.org/wiki/Russian_interference_in_the_20...

> Often, these tweets are just a useless acknowledgement of some obscure holiday celebrated by some unrelated ethnic or religious group that they want the votes of

Lol. Yes that's a mandatory part of being a politician. I hardly think that when most people think of the problems of twitter/social media they are thinking of Trudeau tweeting "congrats Mattea Roach on winning 23 consecutive games of Jeopardy" or whatever. Btw, actually I just scrolled through his twitter and his tweets are almost entirely practical https://nitter.it/justintrudeau.

Anyway, the problem here is bot farms creating tens of thousands of sockpuppets posing as "average citizens" to push some sort of narrative and try to manufacture consent, which is qualitatively different from what you are describing.

One of the secrets of getting positive value out of Twitter is to block all or most politician accounts you come across. Mute the names of the popular politician for extra measure.

Exactly this. It takes some work to set up, but the reward is plenty. At least the bare minimum tools are available to enable this. In addition, what other platform even lets you block the advertisers, giving you at least some deciding power over the adds you get.

A politician is an organization, as much as any corporation or brand is. The politician is the owner, spokesperson, and mascot, but the more senior they are the more they have a huge staff that is responsible for what would otherwise be called CEO, CFO, Chief Council, PR, HR, IT, R&D.

And remember that corporations are people too, my friends. But they are very strange kinds of people. When Taco Bell hires a marketing firm to pay a TV station to air a commercial X times, is that a bot? Or even stranger, Taco Bell and Pizza Hut can get into a twitter troll war with each other, even though they're both wholly-owned subsidiaries of Yum! Brands, Inc. and the whole thing could have been written by a marketing firm and scheduled it through TweetDeck. Is that a bot?

Edit: Or every newspaper is basically auto-posting their recent articles, some even just doing it like an RSS feed, where it tweets a link and headline whenever an article is posted. Is that a bot?

Absolutely, yes to all of that.

(Off topic, but I love the juxtaposition of your handle and the topic being discussed. :) )

> corporations are people too, my friends. But they are very strange kinds of people.

It has been said elsewhere that corporations are a sort of amoral artificial intelligence. The trouble is that they are more powerful than a human citizen, while having none of the community or political interests of a human citizen.

The bigger one is the lack of any consequences whatsoever. For instance, do something that leads to the death of another person, even unintentionally, and you may find yourself behind bars for some significant period of time. Do it as a corporation, even on a large scale, and you'll generally suffer little more than a fine that is generally no more than a few days of yearly revenue, and of course zero consequences for the executives who in many cases are directly culpable - frequently making decisions consciously prioritizing short-term profit over all, including safety. Boeing being the contemporary poster child.

> every now and then I check what some elected Canadian politicians have tweeted recently

there's yer problem, eh?

In recent years we've seen what it's like when politicians stop being boring on Twitter.

There are worse things in life than polite artifice. Maintaining it is, itself, an achievement.

What about people who use the API for accessibility reasons because the official app isn't compatible with their needs?

When I used twitter a decade ago, I used TweetDeck, which used the API to post to twitter for me. So that would label me as a Bot just because I like a better UX/UI.

That’s a trivial problem to solve.

So trivial that nobody has ever solved it at scale.

If there's a flow that a human can go through to oauth with an application, then there's a way to bypass that same flow through a bot.

How is that not a problem for the official interface then?

well, it is. so the problem is the quest to get rid of bots.

Let me solve it as a technical person. Whether management wants it or not is not my forte.

You want an API because other apps are better. Let those apps partner with Twitter to allow only humans and not a second order API. If they do that, they get kicked out faster than Fortnite got kicked out. Both Twitter and the API user need to own the relationship and guarantee its sanctity so it will likely require an investment bigger than $99 to be paid by the API user (app) to Twitter.

So you will not have infinite API apps, just a few of them, those who can pay the cost of providing and validating the API.

> You want an API because other apps are better. Let those apps partner with Twitter to allow only humans and not a second order API

I want to write a blog engine that posts a summary and a link to my twitter timeline automatically.

Do I have to "partner with Twitter"? How will that look?

No problem with such account being marked as Bot, right?

Twitter owns TweetDeck now, so that wouldnt be the case today

Telegram does something like this.

But it also allows (even encourages) third party clients, which can act on behalf of ordinary users.

Good example:

AwareOps tweets every time a solution goes down: https://twitter.com/awareops

Is that spamming bot stuff or useful or both?

How about the kid who wrote an app that tweets everyone position of a billionaire's airplane location?

That's the point.

> but if you tweet through the API then you get a big "BOT" tag next to your username on the feed.

What if I do this through Zapier? What about Hootsuite? Who's a BOT and who isn't?

How about we just print capital letter the "BOT" on automated accounts? And then just add a optional filter for comments/likes/shares?

That’s not the bot issue everyone is talking about. It’s bunch of fake/empty accounts posting/replying something what pushes some fakenews, propoganda, etc.

It has been especially noticeable this year, since February 24th

Musk hates bots mainly because of all the єɭ๏ภ ๓ยรк wants to give you free money scams. That isn't really caused by the API

The 'bots on Twitter' thing is perhaps one of the few things I agree with Musk about. Searching a hashtag like #javascript or #webdev or similar reveals a whole universe of obviously low quality content posting bots getting suspiciously the same number of likes and retweets on every post. Things like "Top 10 Web Languages You Need To Know" and the list is just randomly shuffled with link shorteners everywhere. This stuff is definitely polluting whatever sacred public square Musk seems to think Twitter needs to be. Obviously bots and spam is a monumental problem that transcends Twitter.

They added this a year ago:

https://blog.twitter.com/common-thread/en/topics/stories/202...

I have not seen many obvious bot accounts using it, even the “good” ones.

They should try to detect bots not using the API and ban them.

IMO all modern freedom of speech issues stem from not being able to differentiate between humans and bots. And it may be the case that it's not a solvable problem and we should assume any sentiment expressed online is always just noise.

> the problem is the lack of differentiation between bots and real accounts

Nah, I'd bet that it really is as simple as the fact that the first thing Elon sees under any and everything he tweets is a flood of cryptobot fake posts and scams. That he and we identify them quite easily, and he's just flat-out tired of it.

If clamping down on API access also curtails a significant proportion of those posts, then he'll do it, even if it's just a temporary relief.

I agree this is what his issue is likely to be.

I'd be pretty surprised if those accounts were obviously using the API, too easy to ban them by application.

half of what I tweet could be done by a bot, that is to say answer to writerslift posts with links to books and articles. the other half could not. I'm currently real, but would I be a bot if I automate my bot half?

>The problem isn't the bots,

I would see bots as a major concern if you want to monetize the platform.

this is brilliant

That already exists (see e.g. that celebrity jet tracker) but I've never found out how to get that label.