Artificial intelligence and democracy

Democracy has not always been widely accepted in all societies and at all times. Democratic systems have emerged at distinct times and in different places throughout history, suggesting that certain individual and societal conditions are necessary for democracy to exist. Currently, artificial intelligence systems, which significantly impact behaviors, habits, and intellectual structures, are implemented in several aspects of individual and societal life. As a result, there is a high possibility that artificial intelligence has an influence on the individual and societal conditions that are necessary for democratic governance. This study examines the impact of artificial intelligence on democracy. Our study's main focus is on understanding why and how artificial intelligence affects democracy. As a result, the question arises on how to legally address the effects of artificial intelligence. The issue was analyzed within the context of constitutional law using the following question: Do constitutional law's historical mechanisms, including human rights, the rule of law, and the separation of powers, adequately protect democracy against contemporary AI threats? This analysis is limited to human rights within the scope of our study. Artificial intelligence is, from a narrow perspective, a contemporary, dynamic, and highly technical field. Therefore, a descriptive approach, rather than a normative one, has been followed in this study. The influence of artificial intelligence systems, utilized for specific tasks in narrow sectors, on democratic institutions has been examined and the implications of these findings on democratic values have been assessed. We have refrained from analyzing the impact of AI on the democracy of a specific country. While our study is primarily grounded in the fundamental concepts and academic studies of public law, we have also drawn on concepts and academic research from the social sciences, human sciences, and to a limited degree, computer sciences. Our thesis consists of an introduction, three main chapters and a conclusion. The first chapter provides an in-depth analysis of artificial intelligence, covering its basic elements, functioning, general characteristics, technical applications and areas of use. The objective of this examination is to establish AI's conceptual framework. Artificial intelligence is a broad term that encompasses both traditional artificial intelligence and machine learning methods. In the early years of modern AI research, programmers primarily employed traditional programming-based AI methodology. In this approach, the programmer first identifies a goal that requires intelligence, determines the necessary steps to achieve it, and then manually instructs the machines to perform each step in sequence. Machine learning, a sub-branch of artificial intelligence, involves the analysis of large amounts of data to uncover underlying patterns. In this method, the machine learns on its own by analyzing millions of data points related to the action considered to require intelligence. Therefore, "big data" becomes a vital resource for this method. As machine learning is currently the predominant approach in artificial intelligence research, our thesis is based on machine learning method. In machine learning methods, various types of learning are used, including supervised, unsupervised, reinforcement, and deep learning. The machine learning process varies depending on the type of machine learning employed. To perform machine learning, the process involves data preparation, model selection, training and testing, and deployment. One of the most important features of machine learning is its capability of using analyzed training data to predict outputs for unseen inputs. This feature is known as 'generalization'. The 'generalization' feature of machine learning assumes that the model will generate an objective and general outcome that is applicable to any situation, and that the results obtained previously are still valid for future scenarios. The third section of the first chapter the usage areas of artificial intelligence are examined. Artificial intelligence systems are being used in commercial and decision- making applications, promising increased efficiency, cost savings, and higher revenue. The third section of the first chapter provides a thorough examination of the areas of use of artificial intelligence. Given their integration in essential areas, it is inevitable that artificial intelligence systems will have an impact on personal and social life. The fourth section of the first chapter of our study, we reached a definition of "artificial intelligence in a broad sense", facilitating a comprehensive understanding of artificial intelligence's impact on democracy and paving the way for effective legal regulations. Viewed from a narrow perspective, artificial intelligence is defined only as a technical system. However, such a narrow view would reduce the influence of artificial intelligence on democracy to a mere technical problem, with solutions sought through technical enhancements. Hence, within the scope of our study, artificial intelligence is defined as a "socio-technical system possessing a significant industrial infrastructure alongside its technical characteristics." The second chapter of our study examines the impact of AI on democracy. The first section of this chapter presents the conceptual framework of democracy upon which our research is founded, as well as the social conditions that promote democracy. The meaning of democracy can vary depending on context, interpretation, theories and methodologies adopted. Our study is based on liberal democracy. Our study investigates the impact of technical and socio-technical systems on existing democracies, so the concept of democracy is addressed both in terms of its ideals and from an empirical perspective. The ideals of democracy comprise full participation and equality. An empirical comprehension of democracy includes an approach that examines democratic states and uncovers the principles and institutions of democracy. Effective participation, equal voting rights, information access, control over political agendas, and adult involvement are established democratic principles in the empirical democratic framework. Moreover, the democratic institutions, defined in the empirical democratic framework, encompass free and fair periodic elections, citizen participation, freedom of speech, access to alternative information sources, and institutional autonomy. Favorable social conditions for democracy involve democratic beliefs and political culture, a modern market economy, and a pluralistic society that remains unfragmented. In the second chapter of our study, we examine the effects of AI on democratic participation. In liberal democracies, elections, representation, and public opinion form the vehicles of citizens' democratic participation. Citizens exert control over political authority through free, fair, and frequent elections. The presence of free and fair elections first and foremost necessitates a democratic public opinion. The principles and institutions defined by empirical democracy are the tools that shape and safeguard a democratic public opinion. A democratic public opinion can only form within a democratic public space where all citizens can participate and freely debate public matters. Today, people and things are increasingly networked, and a significant part of their lives takes place in cyberspace. This makes cyberspace a new platform for democratic participation. According to Habermas, a fresh mode of public communication has arisen in this environment, brought about by recommendation systems, a type of AI system. These systems scrutinize previous online behavior to create unique personal profiles for each user. In these personalized cyberspaces, individuals often encounter previously come across ideas and people, leading to the formation of filter bubbles and echo chambers. The possibility of encountering random or different points of view in this environment is significantly reduced. Semi-public cyberspaces, where individuals create their own public issues, are 'fragmented realities'. People's thoughts become more extreme every time they hear their own voice louder. Thus, cyberspaces are highly conducive to deepening social divisions and fostering group polarization. The second section of the second chapter of our study examines the barriers to democratic participation in cyberspace. AI-enhanced cyberspaces facilitate the external interference in shaping public opinions on societal issues. The Council of Europe has identified the capability of AI-enabled systems to anticipate and manipulate people's actions in social, economic, as well as political contexts. Customized profiles offer ample opportunity for manipulation via micro-targeting. Microtargeting is the practice of reaching possible voters with the intention of influencing their opinions and electoral decision-making. Several microtargeting techniques exist depending on the personal data used and the personal profiles created. Fake information, fake images, deep fakes, computational propaganda with bots and troll accounts created by AI-enabled systems distort the information element of public issues in cyberspace. Media organizations' decreasing ability to verify news accuracy in such environments also undermine trust in cyberspace. Online content monitoring activities carried out by the companies behind social media platforms pose another obstacle to democratic participation in cyberspace. Surveillance, conducted autonomously or semi-autonomously for commercial purposes, often results in the removal of content or accounts that violate the platform's rules. Therefore, it may be stated that social media companies now play the role of gatekeepers in the networked society, which once allowed mass communication tools to control the agenda. The third section of the second chapter of our study examines the impact of AI on equality. First, we introduce the conceptual framework of equality. In implementing the principles and institutions of democracy, the study acknowledges the adoption of the political and legal equality criteria by liberal democracy and includes a minimum criterion of social equality. This criterion suggests that there should not be significant economic and social disparities between social groups. The sociotechnical nature of artificial intelligence, which reflects human opinions and value judgements, is likely to reflect human cultural and historical biases. As artificial intelligence systems become more widespread and generalized, bias- influenced discriminatory outcomes are becoming the norm globally. Moreover, the unclear functioning of AI systems, combined with the perception that these systems make impartial and unbiased decisions, worsen their adverse effects. The effects of AI on equality can be divided into two categories: 'allocation harms' and 'representation harms'. The use of biased and discriminatory AI systems in criminal justice can impede equitable treatment under the law. Similarly, if used in the allocation of opportunities and resources, AI systems can worsen economic inequalities. These situations fall into the category of 'allocation harms'. Representation harms refer to damages that reinforce the exclusivity of a group and deepen social stratification. Representation harms, particularly those that persist over time, shape people's perceptions and opinions, thus impeding the development of a democratic political culture. The consequences of AI systems perpetuating stereotypes, failing to recognize a societal group, using derogatory language or causing underrepresentation of certain groups, can be cited as examples of representation harm. The policy of classification, which is inherent and persistent in the functioning of machine learning, constitutes the central impact of AI on equality. Classification is the process of categorizing individuals or objects using predetermined labels. It is probable that discriminating between different groups based on their categorization may exacerbate societal inequalities. Individuals are regarded as members of the group into which they are classified. To classify, it is necessary to ascertain both the basis and the criteria for classification. Thus, in the classification process, it is essential to accord priority to specific values and contexts. Indeed, by its very nature, classification imposes a social order that reinforces existing hierarchies and can intensify societal inequalities. The third section of our study's second chapter examines the impact of AI on political power. It comprehensively evaluates the concept of political power, its connection with surveillance, the changes in surveillance due to AI, and the digitalization and consolidation of political power. Political power is defined as the obligatory relationship between law and enforcement. Although enforcement entails the use of coercive force, political power is made legitimate by the superior principle of law. Social and political shifts over the course of time have given rise to the modern state, a form of political power that is institutionalized and differentiated. In a modern state, the nation makes laws, but the exclusive legitimate owner of coercive force is the modern state's apparatus. Liberal democracy has also developed within the framework of the modern state. Artificial intelligence systems, with their unique qualities and critical areas of application, deepen surveillance, a fundamental tool of political power. This not only reinforces modern state structures but also leads to private actors challenging the modern state. Therefore, our study examines whether it's possible to consider these private actors as a new type of political power. The influence of AI on political power requires a reassessment of the efficiency of constitutional measures that safeguard liberal democracy. The third chapter of our study analyses whether the human rights are enough to protect democracy against AI's threats. Human rights are constitutional norms that ensure the realization of the principles and institutions of democracy. Therefore, it is crucial to protect human rights against the threats of artificial intelligence to democracy. In the second section of the third chapter of our study, current artificial intelligence regulations are analyzed regarding their impact on democracy and human rights. This review is carried out through the regulations of the USA, China, the European Union, and the Council of Europe. The European Union's approach to the risks of artificial intelligence is to promote human rights and democratic values in artificial intelligence within a regulatory framework. Therefore, the EU plays a pioneering role.