Further Readings

Further Readings on AI in general

Further readings on Populism and AI

“Populism has been defined in many different ways, mostly in regard to political ideology and political dynamics, but only in recent years in relation to communication variables. The aim of this paper is to contribute to the identification of a socially mediated type of populist communication profoundly affected by the specific nature of social media. It presents and discusses empirical evidence on Italy’s populist and non-populist leaders that use Facebook regularly, and highlights the extent of the overflow of populist communication patterns and ideological features into mainstream political communication. Populist ideology fragments emerged in Italian leaders’ Facebook posts, thus leading to two main conclusions: first, populism appears to be ‘endemic’ in the Italian online facebooksphere; second, political actors—even non-populist ones—do not disdain the adoption of typical populist rhetorics.” (Quote by Gianpietro Mazzoleni & Roberta Bracciale)

Chantal Mouffe and Populism

  • Snarky Guardian review of her new book
  • The key characteristic of all populism, Mouffe writes, is the identification of a “people” who are distinguished from some kind of adversary, a distinction that serves to unite and mobilise them. Nationalists can point to any number of adversaries, from foreign powers to immigrants to “enemies within” (the liberal media, socialist intellectuals, Jews), all of whom can be charged with harming “the people”
  • “Margaret Thatcher succeeded in building just such a hegemony, painting her policies as the only way of acting in the nation’s interests, such that a common-sense view of the economy was already in place by the time New Labour came to power”
  • 2005 book on Latin American populism by lifelong collaborator and co-author Ernesto Laclau (Arg)
  • Populismo de Izquierdas

Misc Journalistic Hot-takes


Online Toolkits and Resources

Fake news, disinformation, and data manipulation

Bot detection materials

Dark ads and Fake News identification tools

For Journalists

Guide for journalists (partial draft)


There is undoubtedly a measurable, worldwide rise in the violence and menace that rightwing extremism and fascism are inflicting on marginalized and vulnerable populations. Simultaneously, these voices, ideas, and actions are being normalized and “mainstreamed” in many of the world’s political forums and cultural conversations, largely because of shifts in the management, staffing, writing practices, and editorial policies of conventional journalistic institutions, and largely because most of the world (and its understaffed news rooms) increasingly gets its news, opinions, and the concepts that frame conversations from non-journalistic sources delivered by algorithmic social media. In many ways, these algorithms (and the entirely human decisions of the corporations in charge of them) seem to favor this shift away from classical journalism and towards new forms of media that the farright, populist, and/or anti-democratic forces seem more successful at appropriating. Many observers of these developments harbor a growing sense of uncertainty as tech giants continue to enforce community standards in a way that silences marginalized groups and favors extremists- all while responding incredibly slowly to the growing reality of offline violence spurred by online speech, from the USA to Germany to Sri Lanka.

We want to highlight the stories of people or communities that have been negatively impacted by the manipulation of AI and algorithms of new-media/social-media platforms by right-wing interests. We hope that these case studies will make clear the pitfalls and ethical obligations of journalism in our new age of algorithmic decisions about what content gets seen, or is even allowed to exist online, helping all journalists and information workers to come together with a common framework for understanding and action. We have assembled some resources:

Case Histories

Examples of Communities Being Impacted, sorted by category

Filter Bubbles, Fake News, Promoted Hateful content

US Case-studies and technical analyses

Electoral Interference

Facial Recognition

  • Persecution of political opponents in the Ryazan region in Russia, the Ministry of Internal Affairs uses a mobile biometric system equipped with AI face recognition. This allows them to identify individuals at public events, which increases the likelihood that the police will be able to use this technology to identify and arrest political protesters.

Customs and Immigration Canada

  • Bots at the Gate
  • Less definitive examples of right-wing co-opting/playing mainstream news and/or algos Border hysteria 2018 (caravan nonsense) versus 90s border hysteria (pete wilson, minutemen, TV-news-based campaign to demonize “brown wave”

Keyword Wordplay

How does right-wing populism operate linguistically and how might this provide a window into the curious ability of this discourse to spread affect in viral ways? Re-coding Populism participants collected piles of bile and turned them into algorithmic poetry, presented as a handmade “zine” which can be reproduced anywhere by printing and binding the following PDF.

Algorithm Analysis

YouTube’s recommendation algorithms are a black box – we pump our preferences in and it provides us with suggestions of more things to watch, but it is impossible to know how this works. In this experiment, two participants created new “clean” Google accounts, submitted identical YouTube searches and clicked on the same recommended videos to see at what point their journeys diverge.


Glossary of Artificial Intelligence, Populism, and New Media

Artificial Intelligence and Information Technology core concepts

  • Algorithms: a list of steps to follow in order to solve a problem: most relevant for AI-related work is algorithms written in code that tell computers how to do something; Very important to note the difference between an ordinary algorithm, which only completes the task it was programmed for, and a machine-learning algorithm, which can generate new data, and grow with that.
  • Artificial intelligence: there is lack of a generally agreed upon definition, but AI is usually used to refer to getting machines to think on a ‘deep’ level, modifying and refining their own programming over time (after processing huge data sets and “feedback loops”) rather than simply running through static routines and programs over time. The most basic technical definition of AI would be a self-refining program structured as a massive collection of Bayesian trees, mapping all possible outputs and choosing between them probabilistically. (DK)
  • Machine learning: “Machine Learning is the science of getting computers to learn and act like humans do, and improve their learning over time in autonomous fashion, by feeding them data and information in the form of observations and real-world interactions. (Source); Machine learning is a subset of Artificial intelligence. The terms machine learning, algorithms, and artificial intelligence cannot be used interchangeably
  • Machine learning algorithms: algorithms that use principles of machine-learning to learn from data and come to an answer without having explicit instructions
  • Bots — Algorithms impersonating humans, similar but distinct from “trolls” who are (often paid) actors impersonating other humans
  • troll/bot farms: An organization or company whose employees or members attempt to create conflict and disruption in an online community by posting deliberately inflammatory or provocative comments. These farms can be hired to generate online conflict or trigger flame wars. A common strategy deployed is to create a concern troll, a false flag pseudonym created by a user whose actual point of view is opposed to the one that the troll claims to hold. The concern troll posts in Web forums devoted to its declared point of view and attempts to sway the group’s actions or opinions while claiming to share their goals, but with professed “concerns”. The goal is to sow fear, uncertainty and doubt within the group
  • Data Analytics (sometimes referred to as “Data Science”) – the complex toolkit for making analysis and decisions on the basis of huge data sets (“Big Data“), whether for academic, governmental, or private-sector purposes. Increasingly Data Scientists are becoming more central to the executive decision-making of corporations and governments across industries and contexts. Traditionally, Data Analytics (in any context) are only as useful or as neutral or as reliable as their data sets and its provenance: data accrued over decades in an institutional context generally will bestow analyses based on them with the biases and weaknesses not only of decades of data-collecting but also of that institution, its staffing, its time and place…
  • Concrete example: “So it used to be that life insurance companies made black men pay more for life insurance than white men simply because they were going to die sooner. That lasted for a long time before the regulators in question were like – wait a second – that’s racist.” (Source)
  • Some data sets are “harvested” from the main institutions of the “freemium” internet (free services or free-for-most-users services that analyze huge amount of traffic) – others, like geographical information, are largely crowdsourced, while others are cobbled together from commercially-available data like credit histories, shopping habit profiles, etc.
  • voter rolls (raw data of most profiling, like Cambridge Analytica) are huge data sets of all registered or voting voters across multiple elections in a given district, which until a few years ago were readily available in many states, often without even being anonymized or having personally-identifying information obscured.

Advanced topics in Artificial Intelligence

  • dark ads, different kinds of paid social media work, including “user-generated” content made by paid human or bot agents
  • viral content, etc (“bought trends”) – see the gaming section below
  • “process laundering” – delegating to an AI or algo the work that you don’t want individual discretion for and potentially liability
  • predictive policing – using big data (simple algos or AI) to flag “likely” criminals, “gang affiliates” (as opposed to members), or other classes of people for differential treatment
    • arguments against “profiling” largely apply– “profiling” is using police discretion
    • predictive policing sometimes seen as “laundered” profiling/discrimination
    • another example is using AI content moderation to monitor and remove content that violates copyright or other policies instead of doing it with human moderators- a use much discussed in terms of Article 13 the EU’s pending Copyright directive

New Media core concepts

  • deepfakes: manipulated images, video, and audio generated by sophisticated applications of machine learning- but also still, at this point, detectable by machine learning and other techniques and tools (for more see here) (DK)
  • Different forms of engagement: affective (“subliminal”, emotional, psychological, rational, ideological): Affect: “Affect arises in the midst of in-between-ness: in the capacities to act and be acted upon. Affect is an impingement or extrusion of a momentary or sometimes more sustained state of relation as well as the passage (and the duration of passage) of forces or intensities. That is, affect is found in those intensities that pass body-to-body (human, non-human, part-body, and otherwise), in those resonances that circulate about, between and sometimes stick to bodies and worlds, and in the very passages or variations between these intensities and resonances themselves”. Gregory J. Seigworth and Melissa Gregg
  • The attention economy – as much of the business model of the Internet for the last twenty years has monetized based on metrics of clicks, page views, and traffic, the metaphor of attention as “currency” in a virtual market for literal or figurative “attention” has taken hold on the media space and its economics.
  • Clickbait is a text or thumbnail link that is designed to entice users to follow that link and read, view, or listen to the linked piece of online content.. Click-bait headlines typically aim to exploit the “curiosity gap”, providing just enough information to make readers of news websites curious, but not enough to satisfy their curiosity without clicking through to the linked content Clickbait feeds on the general lack of attention that defines online strolling
  • Memes: Memes are multimodal artifacts remixed by countless participants, employing popular culture for public commentary
  • virality vs memes: A viral message or object must spread without loosing its original idea or content, memes can help or disrupt virality as they transform and alter the message
  • flame warfare: Flaming is the online act of posting insults, often laced with profanity or other offensive language on social networking sites. These insults may turn into flame wars where two or more individuals unite to exchange or unite against a third party with verbal attacks. (Wikipedia)
  • virtue signaling: is the conspicuous expression of moral values. The term was first used in signalling theory, to describe any behavior that could be used to signal virtue—especially piety among the religious. In recent years, the term has become more commonly used as a pejorative by commentators to criticize what they regard as empty or superficial support of certain political views, and also used within groups to criticize their own members for valuing appearance over action (Wikipedia)
  • superposters: “Superposters tend to be ‘more opinionated, more extreme, more engaged, more everything.’” Their influential role further distorts perception of reality. These posters had a significant impact during the anti-refugee violence in Altena, Germany.

New Media Advanced Topics

  • content moderation: the process whereby a host of content such as videos, blog posts, status updates, etcetera, determines what content can remain on the platform. The term is generally used to refer to commercial content moderation as done by platforms like Facebook and YouTube. For more information, see the film the Cleaners and UN Special Rapporteur Freedom of Expression David Kayes June 2018 report; Note: working full time (or more than full time) in content moderation may cause emotional distress and even PTSD
  • gaming” algorithms – taking strategic advantage of how content is sorted, aggregated, or spread.
  • Search Engine Optimization (SEO), “shareability”, and “trend-friendliness” – all three of these refer to different in the earliest example of this, Google’s search engine ranking mechanisms were relatively transparent, and a whole industry (“SEO”) was spanned re-writing and tweaking online content to get more traffic and attention by optimizing for these algorithms.
  • disinformation – deliberately confusing the public by spreading and popularizing false information to influence their thoughts and action
  • disinformation taxonomy from First Draft new: Seven types of dis and misinformation (it’s important to not use the blanket term “fake news”)
  1. satire/parody,
  2. false connection (when headlines/caption don’t support content),
  3. misleading content,
  4. false context (share with false information, such as saying “this photo is from Syria” when it is from Yemen),
  5. imposter content (impersonating genuine content),
  6. manipulated content (including “doctored” audio or video, selective omissions, etc.)
  7. fabricated content
  • dismediation – openly or subtly attacking mainstream/traditional media and getting the public to lose the distinction between propaganda and media (relies primarily on imposture content) – Source
  • filter bubble: Term derived by Eli Pariser that describes “that personal ecosystem of information that’s been catered by these algorithms to who they think you are…” isolation which result from personalized searches when a website algorithm selectively guesses what information a user would like to see based on information about the user, such as location, past click-behavior and search history. As a result, users become separated from information that disagrees with their viewpoints, effectively isolating them in their own cultural or ideological bubbles.
  • bothsidesism” – the inappropriate framing of a conflict as a difference of opinion or viewpoint between two groups meeting as equals – the most extreme example in recent years has been Trump saying of the Charlottesville demonstration and riots, “You had some very bad people in that group, but you also had people that were very fine people, on both sides,” Trump said. – Source
  • Horseshoe theory – in political science and also popular discourse,[1] the horseshoe theory asserts that the far-left and the far-right, rather than being at opposite and opposing ends of a linear political continuum, closely resemble one another, much like the ends of a horseshoe.

General vocabulary for politics and activism

  • Anselmo’s distinction between “the political” (narrowly political, “mainstream politics” – electoral system, etc) and “politics” (in the broader sense of communities, values, activism, beyond and partially or fully outside of that system) – sometimes called “capital-P” Politics versus “lowercase-p” politics, or formal versus informal politics.
  • populism – a form of politics that works primarily/centrally AGAINST the political altogether (in its most extreme forms– i.e. fascism) or AGAINST the centrist/mainstream institutions and parties of the political– “all politicians are crooks except ours”, cynical rejection of political traditions, journalism and academia as institutions of consensus/common frameworks, etc.
  • populism tends to cooperate with older forms of autocracy or demagoguery to make the invalidation, overriding, or scrapping of due process more palatable or “commonsensical” to the people
  • nationalism – the ideology that the priorities and necessities of the nation (however defined) are at odds with international priorities, or the application of an ostensibly “national” agenda to internal affairs in a way that trumps or rescinds earlier commitments (such as safety nets, economic programs, cultural programs, etc)
  • ethnonationalism or ethnic nationalism – the tendency to define the nation as naturally or necessarily coextensive with an ethnic group (usually relying on dubious race science and/or heavy historical revisionism), which also conveniently creates an ethnic other to be united against and scapegoated. “white nationalism” is often used as a shorthand, but it’s worth mentioning that increasingly ethnonationalist movements have deracialized (at least in their outward communications) much of their messaging to refer to “European civilization” and to include high-profile members of color to blur or complicate the simply racial nature of their ethnic identification and their justifications for hardline nationalism.

The Re-coding Populism workshop at Ambient Revolts brought together a diverse group of 17 to discuss and address the connections between artificial intelligence and right-wing populism. The group developed three projects which explore these complex issues. There was a poem generated from a large, AI-processed corpus of extreme right-wing speech (and other readily-shared sources of charged discourse). Another poll-based AI-generated word-project recombined ideas and phrases sourced from the attendants of the conference. Other members brainstormed and collaboratively wrote some materials for distribution among people in positions of informational or political agency, with an eye to advancing data-based media literacy.

This website was assembled organically and collaboratively by the “Re-Coding Populism” working group at the Ambient Revolts conference organized by the Berliner Gazette at the Zentrum für Kunst und Urbanistik in Moabit, Berlin, Germany. Please credit us if you repost or continue this work elsewhere, and for press inquiries, please contact Juan Caballero or Dia Kayyali. To check out the material the group has created, click on one of the images above. All images on this website are CC-licensed (CC BY 4.0) and were taken by Norman Posselt at the BG conference.

Tactics Guide

Guide to 2018 media tactics for organizers and political activists

Tactics used by the “alt-right”/neo-nazis/populists (pick your favorite term)

  • Memes: Alt-right pages on social media sites, from Reddit to Facebook to Gab, are very fond of using memes to gain followers and spread disinformation. They respond very quickly to current events. The AfD Facebook pages have been so successful that researchers in Germany studied how interacting with the online pages increases offline activity.
  • Videos: Groups like the “Proud Boys” and “Rise Above Movement” in the US create promotional videos for themselves.
  • Misinformation (see the glossary for a taxonomy of misinformation): the alt-right is very fond of mislabeling videos to say that the videos show what they want the video to show, or taking portions of quotes out of context
  • staying calm“: The alt-right has a thread running through their videos of remaining calm, centered, and common-sensical in the face of “crazy feminists, dangerous antifa, etc.” this tactic appears to be very succesful. A perfect example of this is the “change my mind” series from Steve Crowder.
  • Responding on their terms to critiques: Populists and the alt-right are fond of responding to charges that they are racist, sexist, violent, etcetera, with a pre-determined set of talking points. They make sure they have a way to respond to every critique. Whether or not their responses are accurate or truthful is not the issue- they simply make sure they’re never caught off-guard by any question or issue that “the other side” will bring up. They will use social media to share responses, and as noted above, they speak in a way that makes them appear reasonable. Here’s a response to the recent arrest and charging of Rise Above Members involved in the Charlottesville protests.
  • Trolling: populists coordinate online attacks, oftentimes specifically with the intent of doxxing (see below)
    • Doxxing: Doxxing means making someone’s private information public in order to target them for physical attacks, threatening phone calls, etcetera- or even just to inflict psychological harm. Here’s an example from the legal team that has been supporting anti-fascist protestors in San Francisco, and here’s an example from the man who captured and shared a the video of Heather Heyer getting run over in the violent 2017 Charlottesville “Unite the Right” march.
  • Astroturfing: Populists, with the help of repressive governments like Russia, will mask planned, co-ordinated, and/or “paid” social media efforts as organic, “grass-roots” ones:
  • uniting around hatred or disdain for an othered group (refugees, muslims, transgender people)
    • dogwhistles: using euphemisms, secret codes (like marking the names of jews or purported crypto-jews in (((triple parentheses))) to signal this status to the knowing reader) and seemingly innocuous details like the numbers 88 or 14.

Tactics for “left” organizers to adopt themselves

  • Global coordination: this is something that has happened to a small extent (including via social media). Examples include Black Lives Matter activists visiting Palestine, activists from favelas in Rio talking to CopWatch.
  • smart spending – The left needs to determine which kinds of social media are more effective than traditional media dollar-for-dollar (or if you’re not spending money, minute-for-minute). Facebook is a place where the “echo-chamber” effect is particularly strong. Considering how to engage elsewhere, or engage with new people, is imperative.
  • ways to “opt out” – fighting fire with water, fighting advertising with community, word-of-mouth, etc
  • memes: the left in many contexts has not adopted a strategy of monitoring and quickly picking up on new memes and deploying them strategically. The left often does not employ targeted/public-appropriate humor or sarcasm in memes, leading to unsuccessful attempts at virality.
  • responding explicitly and calmly to critiques: the left is often defensive in response to critiques about identity politics, antifa violence, etcetera. The left should consider following the “act with the confidence of a mediocre white man” advice. That means responding calmly by both explaining your position but also with the assumption that you are correct. This doesn’t mean that the left should engage in tone-policing of marginalized groups like trans people or black women. It does mean that members of the left with particular kinds of privilege- such as cisgender white men and women, especially those who aren’t poor or working class- should step back and consider whether taking the time to craft a logical response and then engaging calmly and logically is a meaningful use of their privilege.
  • WhatsApp campaigns: the left in Brazil did attempt to adopt this tactic as well, but was not able to compete with the mass budget purchased in packages by businessmen in support of Bolsonaro. However, as WhatsApp continues to gain popularity globally, this tactic should be explored as a tool of community-organizing and counter-information

Marketing/communications techniques that a civil society or other larger organization might consider adopting

  • Consult a technologist who understands these technologies and be clear for yourselves and honest with the people you reach out to about the privacy implications of any tool or technique you are using.
  • Identify your potential target: young people, voters, activists.
  • Find community leaders (people marketers would call “influencers”) and try to work with them or their audiences
  • Create a campaign for every large, motivated community. Amplify their concerns, power up their activism, put them in contact, praise them.
  • Create Memes that are topical and timely; use tools like “know your meme,” keep track of who the top pop artists are, and be funny!
  • Create secret mass groups: Twitter DM, Facebook groups (US Midterms), WhatsApp groups (Brasil)
  • Find the right platform: adults whisperer–> Facebook; newsroom whisperer –> Twitter; Masses manipulation –> Youtube
    • example: for an NGO like WITNESS, Twitter is a good way to connect with peers in civil society and individual supporters. Facebook is the best way to connect with our partners in Brazil. Direct emails are the best way to connect with most funders.
More resources: