Suchergebnisse

The algorithms are formidable governance tools for administrations and now seem inevitable. We would like to show here, through the example of Parcoursup, that they could also become powerful democratic tools. Last January, the government decided to respond to the strong challenge to the PDB algorithm through the hasty implementation of Parcoursup. The discontent with PDB had finally crystallised around the lack of transparency. Unsurprisingly, Parcoursup suffers from the same criticism, multiplied by the fact that each institution now controls part of the process. In a gallery published in the World in December 2016, the academic group Algocit to which we belong took a position in favour of the transparency of PDB and in particular the transparency of its specifications. Our position with regard to Parcoursup remains of course unchanged and we want to clarify this position and the consequences to be drawn for the construction of these algorithms. The replacement of Parcoursup for PDB will make it impossible to avoid a reflection on the procedures for constructing the algorithms used by the administrations. And in the era of the digital republic, this reflection must take the form of a public debate. There are many questions. By whom should the algorithms used by the administrations be developed? How should they be developed? Can the principles governing their construction be defined? How they work? In particular, should they be made transparent? What form could this transparency take? The intention here is not to answer all these questions but to shed light on the background to the debate. It seems to us necessary to show that the answers that will be proposed will depend on the concept of democracy. ; International audience ; The algorithms are formidable governance tools for administrations and now seem inevitable. We would like to show here, through the example of Parcoursup, that they could also become powerful democratic tools. Last January, the government decided to respond to the strong challenge to the ...

Algorithms play a fundamental role in the digital economy. Their impact on the situation of market participants is significant. Hence, ensuring transparency of algorithms, through access to them, is crucial for the proper functioning of the market. Several models of algorithmic transparency are analyzed in the paper: from lack of transparency to complete regulation of algorithms. In particular, transparency through explanation, and "on-demand transparency" were proposed. The goal of the paper is to determine the optimal form and scope of regulation of this area, in order to ensure sustainable competition in the digital market. Hence, the paper focuses on the concept of algorithmic transparency, the nature of the competition in the digital market, the role of algorithms within the digital trade, and problems related to the regulation of algorithms. This allows to answering the question of whether algorithmic transparency is an indispensable condition for sustainable competition in the digital market, and what are the legal challenges, which may arise with respect to various models of algorithm transparency. The paper is embedded within the EU legal framework, discusses new legislative developments in the EU law, such as the proposal for the Digital Markets Act, and includes analysis of EU antitrust case-law and market practices.

AbstractAI systems are quickly being adopted in radiology and, in general, in healthcare. A myriad of systems is being proposed and developed on a daily basis for high-stake decisions that can lead to unwelcome and negative consequences. AI systems trained under the supervised learning paradigm greatly depend on the quality and amount of data used to develop them. Nevertheless, barriers in data collection and sharing limit the data accessibility and potential ethical challenges might arise due to them leading, for instance, to systems that do not offer equity in their decisions and discriminate against certain patient populations or that are vulnerable to appropriation of intellectual property, among others. This paper provides an overview of some of the ethical issues both researchers and end-users might meet during data collection and development of AI systems, as well an introduction to the current state of transparency, interpretability and explainability of the systems in radiology applications. Furthermore, we aim to provide a comprehensive summary of currently open questions and identify key issues during the development and deployment of AI systems in healthcare, with a particular focus on the radiology area.

Development of transparency and compliance algorithms (T2.4) to assure that any data processing is inline with the policies and the legislation (T2.1, T2.2), and to provide synthesis of policies for mined and aggregated data based on the user context, on top of the designed transparency framework (T2.3).

The aim of this deliverable is to dig deeper into open technical and research challenges in terms of personal data processing and sharing transparency, and compliance checking. Towards this end, we pay particular attention to: (i) compliance checking over SPECIAL's policies in Chapter 1 ; (ii) how we can leverage existing distributed ledgers in order to cater for the desired data processing and sharing requirements in Chapter 2; (iii) how we can leverage the Big Data Europe engine for semantic data processing on large-scale RDF data in Chapter 3; and (iv) the guarantees in term of non-repudiation that could be provided by existing fair exchange protocols in Chapter 4. This deliverable builds upon technical requirements from D1.7: Policy, transparency and compliance guidelines V2, the SPECIAL policy language which is described in D2.1: Policy Language V1, and the SPECIAL transparency and compliance framework presented in Deliverable D2.7: Transparency Framework V2.

The increasing popularity of data science has resulted in numerous well-publicized cases of bias, injustice, and discrimination. The widespread deployment of "Black box" algorithms that are difficult or impossible to understand and explain, even for their developers, is a primary source of these unanticipated harms, making modern techniques and methods for manipulating large data sets seem sinister, even dangerous. When put in the hands of authoritarian governments, these algorithms have enabled suppression of political dissent and persecution of minorities. To prevent these harms, data scientists everywhere must come to understand how the algorithms that they build and deploy may harm certain groups or be unfair

Do we have a right to transparency when we use content personalization systems? Building on prior work in discrimination detection in data mining, I propose algorithm auditing as a compatible ethical duty for providers of content personalization systems to maintain the transparency of political discourse. I explore barriers to auditing that reveal the practical limitations on the ethical duties of service providers. Content personalization systems can function opaquely and resist auditing. However, the belief that highly complex algorithms, such as bots using machine learning, are incomprehensible to human users should not be an excuse to surrender high quality political discourse. Auditing is recommended as a way to map and redress algorithmic political exclusion in practice. However, the opacity of algorithmic decision making poses a significant challenge to the implementation of auditing.

The volume presents, in a synergistic manner, significant theoretical and practical contributions in the area of social media reputation and authorship measurement, visualization, and modeling. The book justifies and proposes contributions to a future agenda for understanding the requirements for making social media authorship more transparent. Building on work presented in a previous volume of this series, Roles, Trust, and Reputation in Social Media Knowledge Markets, this book discusses new tools, applications, services, and algorithms that are needed for authoring content in a real-time publishing world. These insights may help people who interact and create content through social media better assess their potential for knowledge creation. They may also assist in analyzing audience attitudes, perceptions, and behavior in informal social media or in formal organizational structures. In addition, the volume includes several chapters that analyze the higher order ethical, critical thinking, and philosophical principles that may be used to ground social media authorship. Together, the perspectives presented in this volume help us understand how social media content is created and how its impact can be evaluated. The chapters demonstrate thought leadership through new ways of constructing social media experiences and making traces of social interaction visible. Transparency in Social Media aims to help researchers and practitioners design services, tools, or methods of analysis that encourage a more transparent process of interaction and communication on social media. Knowing who has added what content and with what authority to a specific online social media project can help the user community better understand, evaluate and make decisions and, ultimately, act on the basis of such information

Automated Profiling

AbstractThe recommendation algorithms on social media platforms are hugely impactful, they shape information flow and human connection on an unprecedented scale. Despite growing criticism of the social impact of these algorithms, they are still opaque and transparency is an ongoing challenge. This paper has three contributions: (1) We introduce the concept of sociotechnical transparency. This can be defined as transparency approaches that consider both the technical system, and how it interacts with users and the environment in which it is deployed. We propose sociotechnical approaches will improve the understanding of social media algorithms for policy-makers and the public. (2) We present an approach to sociotechnical transparency using agent-based modelling, which overcomes a number of challenges with existing approaches. This is a novel application of agent-based modelling to provide transparency into how the recommendation algorithm prioritises different curation signals for a topic. (3) This agent-based model has a novel implementation of a multi-objective recommendation algorithm that is calibrated and empirically validated with data collected from X, previously Twitter. We show that agent-based modelling can provide useful insights into how the recommendation algorithm prioritises different curation signals. We can begin to explore whether the priorities of the recommendation algorithm align with what platforms say it is doing and whether they align with what the public want.

Computer algorithms organize and select information across a wide range of applications and industries, from search results to social media. Abuses of power by Internet platforms have led to calls for algorithm transparency and regulation. Algorithms have a particularly problematic history of processing information about race. Yet some analysts have warned that foundational computer algorithms are not useful subjects for ethical or normative analysis due to complexity, secrecy, technical character, or generality. We respond by investigating what it is an analyst needs to know to determine whether the algorithm in a computer system is improper, unethical, or illegal in itself. We argue that an "algorithmic ethics" can analyze a particular published algorithm. We explain the importance of developing a practical algorithmic ethics that addresses virtues, consequences, and norms: We increasingly delegate authority to algorithms, and they are fast becoming obscure but important elements of social structure.

"This book introduces a Responsible AI framework and guides you through processes to apply at each stage of the machine learning (ML) life cycle, from problem definition to deployment, to reduce and mitigate the risks and harms found in artificial intelligence (AI) technologies. AI offers the ability to solve many problems today if implemented correctly and responsibly. This book helps you avoid negative impacts -- that in some cases have caused loss of life -- and develop models that are fair, transparent, safe, secure, and robust. The approach in this book raises your awareness of the missteps that can lead to negative outcomes in AI technologies and provides a Responsible AI framework to deliver responsible and ethical results in ML. It begins with an examination of the foundational elements of responsibility, principles, and data. Next comes guidance on implementation addressing issues such as fairness, transparency, safety, privacy, and robustness. The book helps you think responsibly while building AI and ML models and guides you through practical steps aimed at delivering responsible ML models, datasets, and products for your end users and customers"--