x       xx       x        xxxx
>Who x >Artificial
Speaks? Intelligence, Language,
x and Democracy
xxxx xx x xxx
  x      x     x
x xxxxx
x xxx


This glossary offers markers in the discussion about Artificial Intelligence, Ethics, States, Democracy, Computation, and Technology. It should serve less as a comprehensive explanation of terms one might encounter, and more like indicators to points of interest in the contemporary discussions around AI. The terms are therefore markers for pressures around race, gender, technology, and the social systems that incorporate their workings. AI does not emerge from a pristine space, but rather in the midst of the messy political landscape that surrounds us day to day. AI is thus not a miracle cure that will save us from our faults and mistakes, but rather a confrontation with the pressures, injustices, and possibilities that have been embedded in the social environment for centuries. The terms in this glossary serve therefore as dots that need connecting. How these connections are made is not only a task for the institution of democracy, but also for critical participants, protestors, and liberators.

> Autonomous & intelligent systems (AI/S)

Since there is not a general consensus about what intelligence is, every definition strongly depends on (and reflects) one’s own personal set of beliefs. Generally speaking, an Autonomous and Intelligent System is a cluster of interlinked points performing different singular tasks altogether, which is then able to represent symbolically the reality, compute high amounts of information and react to the external world. That might be either the Roomba vacuum cleaner and the military drone driven by the algorithm capable of recognizing an enemy’s target, but also a termite colony or the network of our neural cells.

> Digital colonialism

A view, according to which data is controlled and owned by global corporations of the Global North, such as Google or Facebook. By extracting user data from a variety of sources, Google extracts the world’s data flow into its corporate cloud. By processing it for consumer and business services it concentrates power and resources in one country – the US. Poorer countries being often overwhelmed by this technology cannot keep up developing their own industries and products to compete with Western corporations. Opposite to it stands data localisation, according to which data should be stored and managed locally.

> Machine-learning

The umbrella term machine learning stands for the scientific study of artificial learning, that means, the process of acquiring new, or modifying existing, knowledge and skills. It can be subdivided into Supervised Machine Learning, Unsupervised Machine Learning and Reinforcement Learning. The term is often distinguished from the subcategory ‘Neural Networks’ and used synonym to symbolic or so called GOFAI. ‘Good Old-Fashioned AI’ builds and processes highly complex statistic models based on trainings data and recognises and finds new patterns in them that allow for predictions.

> Algorithms

Even though the term ‘algorithm’ does not have an absolute or widely accepted definition, it is mostly agreed to be a defined generalized process that is creating an output based on its instructions and input. Algorithms can be represented in a wide range from numbers and symbols to formal or natural language. The latter tends to be ambiguous due to its interpretability and is therefore rarely used for technical complex algorithms. A good algorithm is referred to as a precise and therefore unambiguous set of step-by-step instructions that is finite but can contain iterability – that is, according to Derrida, the repetition of data with alterity.

> Counterfactual explanation

A hypothetical situation that differs from the observed reality, highlighting the causal relation between different events (e.g., if I stopped to breathe, I would die; therefore to survive I have to keep breathing). In the context of Machine Learning, counterfactual explanations are used to give an insight into the way a prediction model works, by feeding inputs to see on which feature a different outcome would be dependent on. But as Machine Learning models rely on pattern recognition instead of direct causality, for each prediction there are multiple counterfactual explanations – and no one is the true one.

> Settler common sense 

Settler common sense refers to a set of dynamics introduced by Mark Rifkin in 2013. Settler colonialism is seen the exertion of control by non-Natives over Native peoples and lands. This takes part in the everyday life of non-Natives by understanding institutionalised framework as extra-political characteristics of humanness. Rifkin reflects on how legal and political structures make it possible for non-Natives to treat Indigenous territories “as given, as simply the unmarked, generic conditions of possibility for occupancy, association, history, and personhood.” Rifkin emphasises that the “terrain non-Natives inhabit and see as given has never ceased to be a site of political struggle”. The struggle is marked by dispossession on the one hand, and the unquestioned assumption – the common sense – that institutionalising approaches trump other modes of living.

> Cyber war

Cyber war is a murky domain that contains in itself several fields such as cybercrime, cyber espionage and cyber activism. The term ‘war’ has been defined differently throughout the years by philosophers and activists. In this context, the cyberwar is concerned with disparate contemporary conflicts between the public and the capitals on the digital environment. The underlying assumption is that the development of information technologies is pioneered as war effort, since the technological revolution is pioneered by governments and military forces. Therefore, every product of information technology, in its forms, structures and functionalities are in favor of control, surveillance and domination. From this perspective, the notion of cyber war stands by the idea that cybernetic capitalism is a product of war itself.

> Homo oeconomicus

The ‘economic man’ is seen from within European philosophy as the figure with an infinite rational capacity who makes decisions based on his own self-interest and happiness. Katherine McKittrick calls him, “the virtuous breadwinner, the stable job holder, the taxpayer, the savvy investor, the master of natural scarcity.” In the field of economics, Homo oeconomicus is used as the ideal decision maker who knows what he wants and has all the information available to him – cutting out the complexity of actual human behavior. In this sense homo oeconomicus is posited as the ‘horizon of humanity’ in the project of colonialism. By elevating a simple model of decision-making, Sylvia Wynter argues that it constitutes a cognitively closed order.

> Cybernetic class war

A term that characterises a condition in which workers performing many manufacturing and office jobs have to compete against the automation of labour involved in digital networks. The cybernetic class war leads to an unemployment of many relatively well-waged mass workers and ultimately drives to a generation of surplus population, meaning those who have been made redundant due to automation of work.

> Racialising assemblages

A theory developed by scholar Alexander G. Weheliye, where race is used as a sociopolitical process of categorizing humanity into full humans, not-quite-humans, and nonhumans. This system depends on the forced embodiment of political hierarchies. In his essay Algo-Ritmo: More-Than-Human Performative Acts and the Racializing Assemblages of Algorithmic Architectures, Ezekiel Dixon-Román expands on this theory to include “the racialized, classed, gendered, queered, and disabled shaping and forming of bodies; bodies that are both human and more-than-human.” Algorithms function as racialising assemblages by targeting non-white agents as recipients of surveillance, violence, and economic dispossession.

> Fake News

A term that became popular and be criticized in 2017, partly because of that Donald Trumps used the term, regardless of the truthfulness of the news, to describe the negative press coverage of himself. Actually the term is a neologism that consist of disinformation that does not come from credible sources and is distributed through fake news sites or social media. It is written and published to damage or manipulate an agency, entity, or person. They often use sensationalist headlines to gain financially or politically. Different types of fake news are: satire or parody, propaganda, clickbait, sloppy journalism, misleading headings and biased / slanted news.

> Internet Trolling

Internet trolling or ‘an internet troll’ refers to a person who posts insulting and disruptive messages in online communities in order to provoke emotional responses.

> Deep Learning

Deep learning is part of the wider concept of machine learning, and relies both on a large amount of available data and powerful computers. It is an artificial intelligence based on the way human brains process data, create patterns and make decisions. Thanks to its neural network and non-linear system, a deep learning AI is able to learn from a vast amount of data, even if they are unlabelled or unstructured. Artificial neural nodes connected together are processing data to learn concepts, where each node can use thousands of neurons. The more and the bigger are these layers, the more performant and abstract the learning is. Since the machine processes on very deep levels, not engineered by humans but emerging from the learning procedure, the abstraction is such that it is usually impossible to understand it as humans.

> Search Engine Censoring

The idea of the internet aimed to democratize the flow of information. In parallel, authoritarian regimes for their interests have curtailed this flow by partially or fully censoring the web. Search engine censoring reflects the power relations between governmental institutions, network operators / providers, and search engine companies. Censorship controlled by governments includes preventing access to information, including IP blocking of foreign websites, political content, or search engine filtering. Users of these search engines often observe evidence of censorship, but the government policies that impose this censorship are not generally public. The policy, mechanism, and extent of this censorship vary from country to country.

> Cyborg

A cyborg is a cybernetic organism, the symbiosis of organism and machine. Using electronics and intelligence to replace and upgrade a specific limitation of a natural being e.g humans, animals and plants. The term Cyborg was first coined in 1960 by scientists Manfred Clynes and Nathan S. Kline in their paper ‘Cyborgs and Space’ suggesting logical reasoning to merge technology and human biology. The most common fantasies of cyborgs are Terminator or Major Mira Killan from Ghost in the Shell, however these fantasies are rapidly becoming reality. Artist and activist for transspecies rights Neil Harbisson is a contemporary cyborg (born with total colour blindness) who has implemented technology into his head that translates colour frequencies into musical notes all through bone conduction for him to understand.

> Crypto-anarchism

Crypto-Anarchism is a cyber-spatial realisation of anarchist ideologies in order to defend and diffuse the individual freedom in the digital era. The idea has been formed and shaped by the Manifesto written by Timothy May, and is still inspiring many movements and activities in the digital open source area. Through the usage of cryptographic software, during the process of sending-receiving info over computer networks, users or associacions have the possibility to hide / reveal their identity, location or any private information. Therefore the application of any laws and regulations of a specific state or country becomes impossible. In addition, the users are given the possibility to create new regulations and laws using smart contracts. Such possibilities create various opportunities in the efforts of providing privacy, political freedom and economic freedom to individuals by freeing them from the surveillance of governmental structures. Crypto-currencies such as Ethereum and Bitcoin, are two examples of such protected contracts, in this case used as a value of money.

> Dataism

The tendency to assume that huge collection of data is equal to knowledge, particularly in the context of social sciences. Seeing everything as a stream of data, dataists believe that the advent of powerful computational machines improved our view of the universe and every complex system: they furthermore foresee an utopia (or a dystopia) in which collecting and sharing Big Data, larger as possible, is the path to progress, whereas human beings, as obsolete algorithms, will yield their decision-making to machines having a better understanding of the world, from fitness app using biometric data to AI-based political systems.

> Surveillance Capitalism

This term, defined by Shoshana Zuboff, refers to the process of making profit based on surveillance and prediction of people behaviors. Companies (Google and Facebook have been pioneers in this form of profit-making) create value from private data by constantly surveilling citizens and consumers data, by storing and analysing them thanks to algorithms and by influencing comportements to make sure the predictions are certainties. One main characteristic of the process is that users conscience, consent and understanding of the system are largely nonexistent, thus raising questions about the asymmetry of power of these companies, the possibility of self-determination and democracy.

> The Singularity

The technological singularity is a moment in time where accelerating development will bring to exponential growth in computation abilities. The results will exceed human intelligence and will radiate outward from the planet until it saturates the universe. The idea introduced by the inventor and futurist Ray Kurzweil, in his book The Singularity Is Near: When Humans Transcend Biology, 2005. The metaphor borrowed from physics, describes what happens when moving through a black hole center to the point of no return.

> Digital Analytics

Mostly used for digital marketing, the term digital analytics refers to collecting, measuring, analysing, visualising and interpreting qualitative and quantitative digital data illustrating user behaviour on the web – and this across different devices. Companies, for instance, will use digital analytics in order to understand, describe and predict the experience of the user and thus improve the overall performance of their websites. Examples of specific applications of digital analytics include SEO (which aims to improve the positioning of individual web documents in search engine result pages), tracking affiliate links, tracking in email campaigns, and many more.

> Psychographics

Psychographics is a methodology widely used in contemporary marketing of studying, collecting and understanding cognitive attributes of consumers – like psychological factors (emotions, values, attitudes) and personal decisions. In contradiction to demographics which is focusing on quantitative data like age, gender or income, psychographics is more dynamic since they are changing like customers lifestyles and values. Psychographic profiles could provide insight into why someone might act, vote and behave on the market.

> Social Engineering

The effort to influence on a large scale human behaviour in societies. Whether done by governments or private groups, there is the underlying idea that, having enough information about human psychology or abundant access to data, a group of people could be controlled as a unique body, exploiting the individual weaknesses, with the aid of propaganda, usage of mass media or microtargeting. The term also refers to the activity of taking advantage of cognitive biases in order to get access to sensitive information (also called human hacking). In both cases, human beings are not seen as independent free agents, but as embedded parts of a bigger system.

> Slacktivism

A combination of slacker and activism. It is normally a pejorative term for minimal-effort engagement to support a cause, with low commitment and risk. Often, slacktivism is practiced through online means or social media by those who are not fully devoted to making a change. Through signing online petitions, posting, liking, and retweeting, the public can easily engage in political movements without much effort. Despite the negative connotation, viral movements fuelled by slacktivism can sometimes have real-world, positive effects – especially if the digital sharing inspires actual action.

> Predictive Policing

A strategy of policing which uses mathematical, analytics and evidence-based intervention models to reduce crime and improve public safety. The data collected from past and present criminal events are fed into an algorithm that runs through complex mathematical and analytical decision making that at the end provides an answer based on the data given. While Law enforcement officials imagine that this gives a chance to foresee and prevent these crimes from happening, the danger of circularity is high by being based on existing data with algorithms functioning to reinforce existing bias. There are multiple methods of predictive-policing and in a study published in 2013 by RAND Corporation they proposed four of these systems including: methods of predicting crimes, predicting offenders, predicting perpetrators’ identities, and for predicting crime victims.

> Robot Tax

A robot tax is a strategy to tax companies that work with robots and Artificial Intelligence. There are multiple ideas on how a robot tax could function. South Korea is the only country that has implemented a form of a robot tax.

Since many governments are mostly reliant on income-taxes, mass automatization will lead to decreased government tax revenue. This money is needed to invest in retraining workers that have been displaced by machines. The concept of a robot tax raises questions about what we consider a robot and on the gap between taxing human labor and automatization.

> Computer vision bias

An inclination towards stereotypical, unfair, based on prejudices or simply wrong predictions  provided by an algorithm trained to describe digital images. In order to achieve the ability of classifying the visual content, a model has to be trained on a large set of images already labeled by humans, such as ImageNet, Open Image or COCO sets, that is: still partial (although huge) collections labeled by a restricted sample of people. This leads to common misinterpretation of objects used in different cultures and especially, regarding face recognition, in reinforcement of gender and race stereotypes.

> Facial recognition engine

Is a system that is capable of identifying or verifying a person by facial textures and shape from a digital image, videoframe or a video source. It is used in airports and border security. They can allow law enforcement to find a wanted person in a crowd or match the image of someone in police custody. These systems are using computer algorithms to compare the specific facial details they gather to a database that contain a collection of data from other faces. This technology also ensures fear, because there are just a few rules on how these systems may be accessed. Studies pointed out that facial recognition may not always be accurate, especially for people of color.

> Weaponized algorithms/technologies

The employment of AI and Machine Learning tools in warfare activities. Those are deployed mostly for logistic activities, speeding up decision-making, increasing situational awareness and – as military organizations say – for locating more precisely enemy targets with less civilian casualties – but, as critics say, relying on the same fallacies algorithms are well known for: biased prediction and unfathomable results. Weaponized algorithms are also present in cyber and information warfare, as enhanced malwares, swarms of bots and fake news generators are already employed for causing instability or division in nation states.

> Interoperability (smart borders)

Interoperability refers to the ways that systems talk to one another and share information. This allows for unrestricted data exchange across different software components even if the programming languages or interfaces are different. Essentially, it refers to the cooperation of information systems. The European Commission created a ‘smart borders’ policy in 2011 to aid border security in the EU that heavily relies on interoperability. Specifically, the four main systems are SIS (Schengen Information System), Eurodac, ETIAS (European Travel Information and Authorisation), and EES (Entry / Exit Systems) – which replaced the manual stamping of passports.