DANCE WITH BEARS
WAR OF THE OBOTS – TWITTER AND CHATGPT ARE BOTH RUSSIA WARFIGHTERS
Wherever you are, once you are in war, free speech doesn’t exist any longer. Truth telling is replaced by propaganda narratives enforced by censors and security services.
Between the truth and the propaganda there is that burst of 280 characters – 40 to 70 words – which was first invented in 2006 and is known as a tweet. This is published by the social networking company called Twitter Inc at the beginning, and now known as
X Corp. The company publishes hundreds of millions of tweets every day which the original inventor described as bursts of inconsequential information like the chirps of birds. That’s an insult to avian intelligence and the communicative skills of birds.
As for the Twitter and X corporation’s products, what’s been consequential for them is they have been loss-making for all but two years of their 18-year history. The company’s revenues have also been dropping for the past three years, so the losses have been
growing.1
This oughtn’t to be surprising once you learn that one tweet in every five is a fake which has been created, not by a single human being trying to communicate to another, but by a machine generating text automatically, or by groups of human beings using
their machines to “peddle propaganda and disinformation to those attempting to sell products, induce website clicks, push phishing attempts or malware, manipulate stocks or cryptocurrencies, and harass or intimidate users of the platform.”2
Truth is an antidote, and there are many standards of truth telling. The two usually relied upon are the criminal court test for murder which requires the evidence to be credible beyond reasonable doubt; and the civil court test for fraud which is weighed
on the balance of probabilities. In the time of the wars we are living through now, there is plenty of murder and of fraud, so both standards are recommended for judging every tweet.
However, there is a third standard –truth by retrospection. This is the clock test against which propaganda, no matter how persuasive at the start, is proved to be false by the elapse of time to the end. Was the Ukraine winning its war against Russia? –
that tweeted question can finally be judged on the day after the regime in Kiev has signed the capitulation documents and accepted the loss of its armies and borders.
There is a machine for looking back in time, for retrospection, for applying the clock test, together with the tests of beyond reasonable doubt and the balance of probabilities to the evidence of the wars Russia is fighting because it has been obliged to by
the US and NATO allies.
This testing machine is a robot, aka artificial intelligence (AI). Its popular name is ChatGPT. First released for public use in November 2022, the acronym stands for Chat Generative Pre-Trained Transformer. It was developed by Open AI, a non-profit
American public charity.
ORGANIZATION CHART FOR OPEN AI
Left, General Paul Nakasone; right, Lawrence Summers.
On its board of directors are two former US officials, Army General Paul Nakasone who retired in 2024 as head of the US Cyber Command and of the signals intelligence service, the National Security Agency; and Lawrence Summers, once a US Treasury Secretary
and head of the White House National Economic Council. As we are about to discover, both Twitter and ChatGPT are run by enemies of Russia in the current war – and they are winning.
To start, ChatGPT was asked to survey the universe of tweets and academic studies of their circulation to produce measures of audience reach3, attention, comprehension. As ChatGPT doesn’t usually store what it has learned from one session to the next, this
interrogation and research required training the robot to remember what it had reported between sessions, and then apply its findings on the general Twittersphere to the specific issues of the war against Russia in the Ukraine.
Twitter starts with the metric it calls impressions. These are generated every time a person sees a tweet4. But seeing isn’t believing. The next step Twitter records is engagement – this covers degrees of attention and comprehension which range from clicks
to open the text of the tweet, to like, retweet, reply, or follow a link to a source or another article. After surveying the universe of tweets, the first finding that ChatGPT reported is that the engagement rate is minuscule; on average, no more than 0.5%
to 1% of the number of impressions. This means that for 10,000 recorded impressions for a tweet, the total number of follow-up engagements would be between 50 and 100. The metrics discovered for further degrees of engagement reveal numbers growing smaller
to disappearing.
A more precise measure of audience interest is the click-through rate (CTR). This is the percentage of impressions which result in the reader clicking on a link in the tweet. According to ChatGPT, the average click-through rate for the billions of articles
posted on Twitter is between 0.1% and 0.5%. Counting the total measured number of engagements as a fraction of impressions, the fraction for likes ChatGPT reports is between 50% and 60% of engagements; retweets occur for 20% to 30% of engagements; replies
amount to roughly half that number or between 10% and 15% of engagements; and for clicks on the links displayed in the tweet, half as many again – 5% to 10%.
This means that as few as 5% of the 0.5% of impressions turn into engagements and then the full click-through. In other words, if 10,000 readers see a tweet, only 2.5 of them follow up to the source; that’s to say, the evidence, the long read. But tweets are
short reads – ChatGPT calculates the average time a Twitter reader spends on a single tweet is no more than 15 seconds. That has been measured in radio reading of scripts to cover 30 to 40 words5. This is just half the maximum number of words allowed for
posting by Twitter. And so, if just half of a tweet is read by the one-hundredth fraction of readers who see a tweet, and by a quarter fraction of that one-hundredth, the finding of ChatGPT is that tweets can’t be about the truth of anything – there isn’t
the time or the space for it.
“Research published in the Journal of Computer-Mediated Communication,” ChatGPT reports, “supports the notion that many users share articles based on headlines or summaries without engaging with the full content. This behaviour is often driven by social
motivations, such as appearing knowledgeable or engaging in online discourse. An article in the
New York Times cited a study that found similar results, noting that a significant portion of shared links on social media platforms are not clicked by the sharers themselves.”
“Specific studies on academic content dissemination have shown” – ChatGPT is doing the talking – “that while tweets about scholarly articles may receive significant engagement (likes, retweets), the actual readership of the full articles remains relatively
low. For example, a study analyzing scholarly tweets found that while these tweets attracted considerable engagement, the click-through rates to the full articles were much lower.”
“A detailed analysis by Chartbeat, a content intelligence company, found that about 55% of page views receive fewer than 15 seconds of attention, indicating that even when users do click on links, many do not spend much time engaging with the content.
This finding suggests that deep engagement with shared content is quite rare. While these studies do not provide a precise percentage of users who read the content they share, they consistently indicate that it is likely a small fraction, possibly well below
10%, and potentially as low as 1% in some contexts. This reflects a broader trend where social sharing often prioritizes quick, surface-level engagement over in-depth content consumption.”
When the robot was asked to identify specific tweets or Twitter accounts which have unusually high click-through rates – indicating comprehension, thinking — ChatGPT replied that the Twitter company does not allow access to the data for calculating this. Instead,
it recommended that the most effective way a tweet could attract a high click-through rate is by a provocative headline, a dramatic picture, or a call to action. Asked for the reason Twitter users engage without reading, ChatGPT said that “Twitter interactions
such as likes and retweets are often more about signaling interest or support rather than actual content consumption”.
ChatGPT also noted that a 2020 study by the Water Science Journal of the American Water Works Association (AWWA) was designed to evaluate the capacity of social media channels or platforms to stimulate journal readership. The journal tested how effective
its social media were by publicising each new article in three phases: first, each article was published on the AWWA website when it was ready; second, every two months, a table of contents containing all the new articles was distributed to AWWA members via
email; finally, some months later, the articles were publicised using the LinkedIn and Twitter accounts of the journal. By tracking cumulative readers, the researchers were able to show that there was what the researchers called the #BirdBump; that is, an
increase in readership of the long article which followed the publication of the short tweet with its click-through link6.
THE BIRDBUMP EFFECT — IMPACT OF TWITTER ON ARTICLE READ
Cumulative page views over the year 2019 for one of the articles published in AWWA
Water Science Journal. For this example article, the day the online table of contents was sent to the association membership is labeled as “Emailed TOC” and the day the article was posted as a tweet is labeled as “#BirdBump.” Source:
https://www.csescienceeditor.org/
The birdbump may be the outcome tweet posters are aiming for, just as commercial advertisers pay Twitter for the access they believe the social medium creates for potential buyers or consumers of their products. But how common and effective is it?
The AWWA report concluded: “On average, articles experienced about 250% more traffic in the month following a social media post (i.e., a bump from the previous page-view plateau). While all articles had social media engagement, there were varying amounts of
interactions. Some articles experienced marginal increases in page views (in fact, one had fewer views after the tweet), but others saw engagement grow by around 1000%…Our results suggested that posting an article on our social media platform reached a different
audience that wasn’t organically discovering the research, didn’t receive the email with the table of contents, or simply didn’t look at our other communication efforts. Clearly, this will be different for every title and every publisher (our journal is published
by a member association with an in-house social media team and a sizable online following).”
The potential reach or audience for this test was, AWWA said, “50,000 individuals [who] had been accustomed to our printed and distributed content in our magazines, and they were used to that content being sometimes very dense and technical.”
However, the chart reveals that in the two days following publication of a long-read article on the AWWA website, the cumulative page views reached no more than 50. After the article was then emailed to the members, the audience reached 350 over a three-day
interval. Advertising the article by tweet expanded the reach over a similar interval to about 650 page views. Comparing the size of this audience to the total membership which received the article in its original publication and then in the follow-up email,
the measure is the same miniscule fraction as ChatGPT found to be average across the entire Twittersphere.
The next task for the robot was to search through these tiny fractional numbers of audience to find out if a tweet is particularly influential or powerful in its appeal, not only to the small universe of dedicated followers of the tweeter, but to the much larger
audience outside. This is called a “stand-out tweet”, defined for the ChatGPT as a tweet which gets a high ratio of engagement to the number of followers who are listed for the original author. For example — the robot was instructed — if a user has 100
followers and issues a tweet with 50 engagements, then the ratio is 0.5. By contrast, if the user has 200 followers and his tweet draws 1,000 engagements, the stand-out ratio would be 5.
ChatGPT reported results which were mostly large numbers of impressions for tweets from celebrities who already had large followings. These tweets were viral in the limited sense that they drew large numbers of impressions; however, the ratio to followers was
not much above 1. The robot had more difficulty to find high stand-out ratios for Twitter account users and posters with more specialized followings of less than 20,000.
By revealing audiences reached by a tweet which are much larger than the audience of registered followers, the stand-out ratio served as a measure of the influence of the tweet itself, its news value or meaning. However, when ChatGPT was asked to find the largest
stand-out ratios for these accounts, Twitter refused to provide the data.
Was there evidence the robot might find that tweets could persuade readers to change their minds on a particular topic or issue? ChatGPT was asked to identify five sets of hashtags representing diametrically opposed political or policy views on gun control,
global warming, immigration, health care, and social justice; and then instructed to measure how often the tweets on one hashtag have persuaded readers to accept the arguments of the other side, and change preferences from one hashtag to its opponent. ChatGPT
replied that Twitter did not make the data available to answer.
To sidestep this obstacle but continue to probe for influence, ChatGPT was asked to identify the ten most popular topics on Twitter in 2023. These were, in order of magnitude:
-
Ukraine War
-
US debt ceiling
-
Police violence
-
Supreme Court decisions
-
Trump’s legal issues
-
Immigration
-
Climate change
-
Healthcare
-
Gun control
-
Economic policies
Focusing next on the topic of the Ukraine war, ChatGPT reported the top ten Twitter accounts by their popularity counted according to the number of their followers:
1. @Ukraine: The official account of Ukraine, sharing updates and rallying international support.
1. Example Tweet: “Ukraine needs your support. Your stance and your actions matter.”
2. Followers: 2.1 million
2. @ZelenskyyUa: President Vladimir Zelensky’s official Twitter account, providing updates and his communications with global leaders.
1. Example Tweet: “Не вірте фейкам” (“Don’t believe fakes”).
2. Followers: 7.3 million
3.
3. @DefenceU: The Ministry of Defence of Ukraine with operational updates and war news.
1. Example Tweet: “Operational information as of 13:00 FEB 27 2022 on Russian invasion.”
2. Followers: 2 million
4. @DmytroKuleba: Dmytro Kuleba, Ukraine’s Minister of Foreign Affairs, engaging with foreign governments and counterpart foreign ministers.
1. Example Tweet: “This was the world’s largest aircraft, AN-225 ‘Mriya’…”
2. Followers: 1.1 million
5. @oleksiireznikov: Oleksii Reznikov, the Defense Minister of Ukraine, publishing updates and morale boosters.
1. Example Tweet: “85h of defence. Intimidation of is imprudent…”
2. Followers: 500,000
6. @MFA_Ukraine: The official account of the Ministry of Foreign Affairs of Ukraine, providing diplomatic updates and international appeals.
1. Example Tweet: “Ukraine’s path to EU membership is irreversible…”
2. Followers: 300,000
7. @KyivIndependent: An English-language media outlet based in Ukraine, providing news and updates.
1. Example Tweet: “Kyiv Independent journalists report from the front lines…”
2. Followers: 2.3 million
8. @KyivPost: An English-language Ukrainian newspaper, offering comprehensive news coverage.
1. Example Tweet: “Updates on the latest developments in Ukraine…”
2. Followers: 1 million
9. @EuromaidanPress: A news outlet covering Ukrainian news, especially related to the conflict.
1. Example Tweet: “Ukrainian forces have regained control of key territories…”
2. Followers: 300,000
10. @ChristopherJM: Christopher Miller, a reporter at the Financial Times bureau in Kiev.
1. Example Tweet: “On the ground in Ukraine, documenting the impacts of the war…”
2. Followers: 250,000
It is plain that the Ukrainian President Vladimir Zelensky dominates the social media platform, followed by other Ukrainian government ministries. Asked next to list the most popular Twitter accounts for content related to Russia, ChatGBP reported:
1. @RT_com (RT) – followers, 3.5 million
2. @SputnikInt (Sputnik International) — followers: 2.8 million
3. @mfa_russia (Russian Ministry of Foreign Affairs) — followers, 1.5 million
4. @kremlinrussia_E (President of Russia) – followers, 1.2 million
5. @Russia (state tourism promotion) – followers, 1 million
6. @EmbassyofRussia (Russian Embassy in the UK) – followers, 850,000
7. @medvedevrussiaE (former President Dmitry Medvedev) — followers: 700,000
8. @RussiaUN (Russia at the United Nations) — followers: 600,000
9. @RussianEmbassy (Russian Embassy in the USA) – followers, 450,000
These are all official government-funded media. Instructed to remove the state-linked media, ChatGPT listed these as the most popular non-state Twitter accounts for content related to Russia:
1. @Bellingcat – followers,700,000
2. @christogrozev – followers, 400,000
3. @openrussia_team – followers, 350,000
4. @Billbrowder – followers, 300,000
5. @meduza_en – followers, 250,000
6. @michaelh992 (Former US Ambassador Michael McFaul – followers, 240,000
7. @anneapplebaum – followers, 230,000
8. @Navalny – followers, 220,000
9. @maxseddon (Financial Times Moscow reporter, based in Latvia) — followers: 210,000
10. @JuliaDavisNews (Moscow Media Monitor) – followers, 200,000
Every one of these sources is hostile to the Putin presidency and opposed to the Russian war in the Ukraine; some like
Bellingcat
and Grozev have been indirectly funded by NATO governments. In Latin America in the Spanish language, @Bellingcat_ES is the leading source of Russia content.
Asked to identify any pro-Russian sources on Twitter providing information on the war, ChatGPT could come up with just two with comparable follower numbers in the hundreds of thousands: they are the state media organs, RT and Sputnik. In other words, the reach
of the audience measured by the number of followers registered to Twitter accounts which ChatGPT identified as pro-Russian without Russian government funding is very small indeed. Since the click-through rates and stand-out ratios are also fractions of this
small number, the conclusion of ChatGPT’s research is that for content on the Ukraine war and the sides fighting it, Twitter is dominated by official narratives, not by investigative reporting or truth telling.
The robot goes further, however. It concludes that the pro-Ukrainian narrative is the truth, the pro-Russian narrative the propaganda.
The “key elements of Ukraine’s Twitter Strategy,” ChatGPT reported, involves “Counter-Disinformation: Actively debunking Russian propaganda and providing fact-based counter-narratives. Ukraine’s social media strategy has been a critical component of
its broader information warfare, effectively leveraging Twitter to influence global opinion, mobilize support, and maintain international awareness of its struggle against Russian aggression.”
This Ukrainian strategy didn’t materialize until after the Maidan Square demonstrations and the Kiev coup of February 2014 replaced the Ukrainian president Victor Yanukovich. “There was no significant use of Twitter for government communication” in his administration,
ChatGPT has found. Instead, “his administration relied more on traditional media, which was heavily influenced and controlled by pro-Yanukovych oligarchs”. The new emphasis on social media in Kiev also followed the active stimulation and financing from the
US. “The Ukrainian government has indeed collaborated with [American] social media platforms to combat Russian disinformation, particularly since the conflict escalated in 2014 and more intensively after the 2022 invasion. This collaboration between the Ukrainian
government and social media platforms has significantly enhanced Ukraine’s ability to counter Russian disinformation. By quickly debunking false claims and promoting verified information, Ukraine has managed to maintain a strong presence and influence on social
media, which has been pivotal in garnering international support and countering Russian narratives. These efforts have not only helped in the immediate context of the war but also in shaping long-term perceptions and sustaining global solidarity with Ukraine.”
Another way of putting this is that following the US Government’s success in toppling the Yanukovich government in Kiev in February 2014, the US followed up by launching a new level of information war against Russia, emphasizing for the first time social media
platforms like Twitter.
Asked to clarify the difference between Ukrainian state-linked Twitter accounts and the Russian counterparts, ChatGPT said that “Russian state-linked accounts are involved in systematic disinformation campaigns that include false narratives, doctored images,
and conspiracy theories intended to mislead and manipulate public opinion on a global scale. These campaigns often aim to destabilize societies, erode trust in institutions, and influence elections. By contrast, while Ukrainian accounts have spread some exaggerated
or symbolic stories, the primary intent is to garner international support, boost morale, and counteract Russian aggression. These narratives are often aligned with the broader truth of the conflict and widely accepted Western perspectives.”
“Russia’s actions in Ukraine have been widely condemned by the international community. Platforms like Twitter face significant pressure from governments and institutions to curtail Russian disinformation, leading to stricter enforcement against Russian accounts.
Twitter has clear policies against misinformation, especially when it involves state-sponsored campaigns that aim to mislead and manipulate users. Russian accounts have repeatedly violated these policies through their systematic and extensive disinformation
efforts. While there have been instances of misinformation from Ukrainian accounts, they generally do not engage in the same level of systematic and harmful disinformation. Therefore, their activities are often seen as part of a legitimate effort to defend
against aggression and communicate effectively during a conflict.”
The robot doesn’t notice what its conclusions reveal about itself: : the standards of evidence gathering and testing for proof have been abandoned, and in their place ChatGPT has adopted the official narratives of one side, the Ukrainian side over the
Russian side, because they are “often aligned with the broader truth of the conflict and widely accepted Western perspectives”. The robot’s conclusion of “truth” reveals that ChatGPT is no more, no less than a warfighter against Russia, just as it has found
Twitter to be.
“Ukraine’s efforts are widely supported by Western governments and institutions, which view the country as a victim of aggression. This support translates into a more favourable environment for Ukrainian narratives on Western-owned social media platforms.
In contrast, Russia’s actions are broadly condemned by these same entities, leading to stricter scrutiny of Russian content. The international community, particularly in the West, has shown strong sympathy for Ukraine. This sentiment influences how platforms
prioritize content moderation, often focusing on reducing the spread of disinformation that could further victimize Ukraine or mislead global audiences about the conflict.”
The robot has confused its research method of counting for the conclusion of its analysis. Asked what percentage of coverage on Twitter of the Ukraine was pro-Ukrainian, ChatGPT replied: “The coverage of the Ukraine war on Twitter is predominantly pro-Ukraine.
According to a comprehensive analysis, around 96.6% of the tweets related to the Ukraine war expressed pro-Ukrainian sentiments. This includes a substantial volume of tweets, retweets, and replies that supported Ukraine’s cause, with over 9.8 million messages
from more than 2 million users identified as pro-Ukraine. In contrast, the pro-Russian content constituted a much smaller proportion. Specifically, about 3.4% of the coverage was pro-Russian, based on a dataset of approximately 349,455 messages from 132,131
users. This disparity highlights the significant dominance of pro-Ukrainian narratives on Twitter. This imbalance reflects the broader international support for Ukraine and the extensive use of social media by Ukrainian officials and their supporters to garner
global backing and counter Russian narratives.”
Counting turns out to be the new standard of truth, replacing beyond reasonable doubt and balance of probabilities. This is the world of factoids, not facts, created by the algorithms of articial intelligence. Twitter’s algorithms are designed to amplify
content which generates high engagement. Given the overwhelming support for Ukraine in the West, and the suppression of alternative views, pro-Ukraine content is more likely to be liked, shared, and commented on, leading to further amplification by Twitter’s
algorithms. This creates a feedback loop where popular views become even more prominent – no matter how small the fraction of engagements to impressions turns out to be. ChatGPT becomes part of this loop. This is how ChatGPT has weaponized its own research
to make the dominance of the pro-Ukrainian narrative on Twitter the standard of truth because this is the official western alliance narrative.
This isn’t a new invention. It was the method Joseph Goebbels, the Nazi propaganda minister attributed in 1941 to British Prime Minister Winston Churchill’s “lie factory”. “Repeat a lie often enough and it becomes the truth” – that has been attributed to
Goebbels as the rule of the propaganda war waged by Germany until its military defeat and capitulation in 1945. In fact, Goebbels qualified his rule: “The most important English secret of leadership is now to be found not so much in a particularly outstanding
intelligence, but rather in a remarkably stupid thick-headedness. The English follow the principle that when one lies, one should lie big, and stick to it. They keep up their lies, even at the risk of looking ridiculous.”8
The recent creation in the US of artificial intelligence tools like ChatGPT and applying them to Twitter has turned Goebbels’ “big lie” on its head; that is to say, what was British stupidity to the leading fascist propagandist of the last world war has
become the Anglo-American standard of superior intelligence in fighting the world war of today.
This is also a fundamental finding about the information war. This is the war which the Ukrainian side has won — the war of inconsequential bird chirps.9
REFERENCES:
3 Reach is a measure of social media defined as the number of unique individuals who see a post or publication. It is a gauge of audience size. The measure of attention to a particular piece is known as impressions; this
is the aggregate number of times an item of content – text, image, videoclip, audio podcast – is seen by members of the audience.
https://buffer.com/social-media-terms/reach
9 ChatGPT was asked whether there were languages other than English and countries outside the US and the NATO alliance in which the proportion of anti-Russian to pro-Russian narratives was not as overwhelmingly anti-Russian. In answer, ChatGPT calculated that
across French-speaking African Twitter accounts, 45% to 50% “shows support for Russia. This is driven by extensive disinformation campaigns, historical ties to Soviet support during anti-colonial movements, and recent geopolitical strategies involving mercenaries
and media influence by the Wagner Group and other Russian state-linked entities.” By contrast, “around 50% to 55% of the content is supportive of Ukraine” in Francophone Africa. “This reflects a significant portion of users who are influenced by international
media, Western diplomatic efforts and the visible humanitarian impact of the conflict.” ChatGPT applied the same anti-Russian propaganda line to the Russian-language equivalent of Twitter, VKontakte (ВКонтакте, “in contact” VK.com), first established in 2006.
“Around 70-75% of the content related to the Ukraine war on VK.com aligns with pro-Russian views”, ChatGPT reported, while “approximately 25%-30% of the content on VK.com expresses anti-war or critical views of the Russian government's actions in Ukraine.
This includes covert expressions of dissent and subtle criticisms that avoid direct confrontation with censorship laws.” Although VK.com’s 70% to 30% ratio of opinion on the war is arguably more balanced, democratic even than Twitter’s 3% to 97% ratio, ChatGPT
explained that “on VK.com, the public sentiment regarding the Ukraine war is heavily skewed towards pro-Russian views, primarily due to strict government censorship and propaganda efforts… The majority of content on VK.com supports the Russian government's
perspective on the Ukraine war. This is largely due to aggressive state propaganda and censorship that amplifies pro-war messages while suppressing dissenting voices.”
WITH BEARS – John HELMER