Reinhard Heil
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Reinhard Heil has a total of 66 co-authors. Here are the top co-authors by number of shared publications:
Christopher Coenen
Karlsruhe Institute of Technology
8 shared publicationsAlexander Unger
5 shared publicationsAndreas Hetzel
5 shared publicationsAndreas Kaminski
5 shared publicationsAndreas Woyke
5 shared publicationsMarc Ziegler
5 shared publications
Latest Publications
- Open Access
- 16 November 2026
- Published by Oekom Publishers GmbH in TATuP - Zeitschrift für Technikfolgenabschätzung in Theorie und Praxis
- p. 1-8
Will AI escape human control?: Expert perspectives on existential risk scenarios of artificial intelligence
The possibility that artificial intelligence (AI) could escape human control and pose an existential risk (X-risk) is increasingly debated in the media and political discourse. While some consider such a threat inevitable, others view it as unfounded or as a distraction from more immediate AI-related risks. To examine these perspectives, we asked German-speaking AI experts to assess loss-of-control scenarios, risk conceptualizations, and discursive assumptions in a Delphi survey and a subsequent discussion. In their opinion, the amplification of existing risks and the gradual delegation of decision making to AI are more likely pathways to a loss of control than AI developing its own autonomous goals. Although most experts take AI X‑risks seriously, they regard them as preventable through governance that safeguards human control and democratic deliberation. Based on the results, we derive recommendations for a precautionary approach to AI X‑risks.
- 16 November 2026
- Published by Oekom Publishers GmbH in TATuP - Zeitschrift für Technikfolgenabschätzung in Theorie und Praxis
- p. 1-8
Open AccessWill AI escape human control?: Expert perspectives on existential risk scenarios of artificial intelligence
The possibility that artificial intelligence (AI) could escape human control and pose an existential risk (X-risk) is increasingly debated in the media and political discourse. While some consider such a threat inevitable, others view it as unfounded or as a distraction from more immediate AI-related risks. To examine these perspectives, we asked German-speaking AI experts to assess loss-of-control scenarios, risk conceptualizations, and discursive assumptions in a Delphi survey and a subsequent discussion. In their opinion, the amplification of existing risks and the gradual delegation of decision making to AI are more likely pathways to a loss of control than AI developing its own autonomous goals. Although most experts take AI X‑risks seriously, they regard them as preventable through governance that safeguards human control and democratic deliberation. Based on the results, we derive recommendations for a precautionary approach to AI X‑risks.
- Open Access
- 9 February 2026
- Published by Springer Nature in AI and Ethics
- Vol. 6 (1) , 152
AI going rogue? An integrative narrative review of the tacit assumptions underlying existential AI-risks
This paper presents an integrative narrative review of the tacit background assumptions underlying AI existential risk (X-risks) futures. Once confined to science fiction, concerns about AI X-risks now shape debates at the crossroads of the tech world, NGOs, politics and (social) media. Despite growing attention, the plausibility of AI surpassing human controllability remains highly contested. Examining 81 peer-reviewed papers from Scopus and Web of Science, we find a fragmented discourse characterized by bold yet often unsubstantiated claims, including accelerationist growth models and speculative calculations of catastrophic tipping points. Anthropomorphic and speculative AI conceptualizations prevail, while interdisciplinary perspectives that consider issues of infrastructure, social agency, Big Tech power position and politics remain scarce. Delineating how these speculative tendencies are detrimental to the current regulatory need to tackle AI harms, we deduce an AI X-risk heuristic and advocate for a shift in attention from the maximum possible negative consequences to the structural and socio-technical characteristics of how AI is embedded—which are the prerequisites for any AI futures to emerge.
- 9 February 2026
- Published by Springer Nature in AI and Ethics
- Vol. 6 (1) , 152
Open AccessAI going rogue? An integrative narrative review of the tacit assumptions underlying existential AI-risks
This paper presents an integrative narrative review of the tacit background assumptions underlying AI existential risk (X-risks) futures. Once confined to science fiction, concerns about AI X-risks now shape debates at the crossroads of the tech world, NGOs, politics and (social) media. Despite growing attention, the plausibility of AI surpassing human controllability remains highly contested. Examining 81 peer-reviewed papers from Scopus and Web of Science, we find a fragmented discourse characterized by bold yet often unsubstantiated claims, including accelerationist growth models and speculative calculations of catastrophic tipping points. Anthropomorphic and speculative AI conceptualizations prevail, while interdisciplinary perspectives that consider issues of infrastructure, social agency, Big Tech power position and politics remain scarce. Delineating how these speculative tendencies are detrimental to the current regulatory need to tackle AI harms, we deduce an AI X-risk heuristic and advocate for a shift in attention from the maximum possible negative consequences to the structural and socio-technical characteristics of how AI is embedded—which are the prerequisites for any AI futures to emerge.
- 1 May 2025
- Published by Springer Nature
- p. 613-615
Negativität
- Open Access
- 4 September 2024
- Published by Wiley in Policy & Internet
- Vol. 17 (2)
Structuring different manifestations of misinformation for better policy development using a decision tree‐based approach
The spread of false and misleading information in digital communication spaces has enormous potential for harm in democratic societies, but has so far been insufficiently addressed by policy makers. This problem has been exacerbated by recent technological developments such as deepfakes. But deepfakes, fake news, and disinformation are only manifestations of misinformation. It is therefore essential to establish a clear definition to develop an appropriate policy. Since misinformation is a hydra whose countless heads cannot be cut off by a single heroine, we conclude that an interdisciplinary approach is necessary to appropriately “regulate” the different dimensions of misinformation. Therefore, we develop a decision tree that allows us to structure the problem of misinformation. Since deepfakes are particularly well suited to characterize the various manifestations of misinformation, we use them as an application example to illustrate our decision‐tree‐based approach. Based on this systematization, it will be possible to identify the capabilities and limitations of strategies from different disciplines and to develop a bundle of measures to address the various manifestations of misinformation. The basic principles of this bundle of measures will already be outlined here.
- 4 September 2024
- Published by Wiley in Policy & Internet
- Vol. 17 (2)
Open AccessStructuring different manifestations of misinformation for better policy development using a decision tree‐based approach
The spread of false and misleading information in digital communication spaces has enormous potential for harm in democratic societies, but has so far been insufficiently addressed by policy makers. This problem has been exacerbated by recent technological developments such as deepfakes. But deepfakes, fake news, and disinformation are only manifestations of misinformation. It is therefore essential to establish a clear definition to develop an appropriate policy. Since misinformation is a hydra whose countless heads cannot be cut off by a single heroine, we conclude that an interdisciplinary approach is necessary to appropriately “regulate” the different dimensions of misinformation. Therefore, we develop a decision tree that allows us to structure the problem of misinformation. Since deepfakes are particularly well suited to characterize the various manifestations of misinformation, we use them as an application example to illustrate our decision‐tree‐based approach. Based on this systematization, it will be possible to identify the capabilities and limitations of strategies from different disciplines and to develop a bundle of measures to address the various manifestations of misinformation. The basic principles of this bundle of measures will already be outlined here.
- Open Access
- 14 June 2024
- Published by Springer Nature in Humanities and Social Sciences Communications
- Vol. 11 (1) , 1-10
What do algorithms explain? The issue of the goals and capabilities of Explainable Artificial Intelligence (XAI)
The increasing ubiquity of machine learning (ML) motivates research on algorithms to “explain” models and their predictions—so-called Explainable Artificial Intelligence (XAI). Despite many publications and discussions, the goals and capabilities of such algorithms are far from being well understood. We argue that this is because of a problematic reasoning scheme in the literature: Such algorithms are said to complement machine learning models with desired capabilities, such as interpretability or explainability. These capabilities are in turn assumed to contribute to a goal, such as trust in a system. But most capabilities lack precise definitions and their relationship to such goals is far from obvious. The result is a reasoning scheme that obfuscates research results and leaves an important question unanswered: What can one expect from XAI algorithms? In this paper, we clarify the modest capabilities of these algorithms from a concrete perspective: that of their users. We show that current algorithms can only answer user questions that can be traced back to the question: “How can one represent an ML model as a simple function that uses interpreted attributes?”. Answering this core question can be trivial, difficult or even impossible, depending on the application. The result of the paper is the identification of two key challenges for XAI research: the approximation and the translation of ML models.
- 14 June 2024
- Published by Springer Nature in Humanities and Social Sciences Communications
- Vol. 11 (1) , 1-10
Open AccessWhat do algorithms explain? The issue of the goals and capabilities of Explainable Artificial Intelligence (XAI)
The increasing ubiquity of machine learning (ML) motivates research on algorithms to “explain” models and their predictions—so-called Explainable Artificial Intelligence (XAI). Despite many publications and discussions, the goals and capabilities of such algorithms are far from being well understood. We argue that this is because of a problematic reasoning scheme in the literature: Such algorithms are said to complement machine learning models with desired capabilities, such as interpretability or explainability. These capabilities are in turn assumed to contribute to a goal, such as trust in a system. But most capabilities lack precise definitions and their relationship to such goals is far from obvious. The result is a reasoning scheme that obfuscates research results and leaves an important question unanswered: What can one expect from XAI algorithms? In this paper, we clarify the modest capabilities of these algorithms from a concrete perspective: that of their users. We show that current algorithms can only answer user questions that can be traced back to the question: “How can one represent an ML model as a simple function that uses interpreted attributes?”. Answering this core question can be trivial, difficult or even impossible, depending on the application. The result of the paper is the identification of two key challenges for XAI research: the approximation and the translation of ML models.
- Open Access
- 15 March 2024
- Published by Oekom Publishers GmbH in TATuP - Zeitschrift für Technikfolgenabschätzung in Theorie und Praxis
- Vol. 33 (1) , 64-67
Künstliche Intelligenz außer Kontrolle?: Interview mit Karl von Wendt
- 15 March 2024
- Published by Oekom Publishers GmbH in TATuP - Zeitschrift für Technikfolgenabschätzung in Theorie und Praxis
- Vol. 33 (1) , 64-67
Open AccessKünstliche Intelligenz außer Kontrolle?: Interview mit Karl von Wendt
- Open Access
- 6 July 2023
- Published by Oekom Publishers GmbH in TATuP - Zeitschrift für Technikfolgenabschätzung in Theorie und Praxis
- Vol. 32 (2) , 58-61
Digitalisierung, Theologie und Technikfolgenabschätzung: Interview mit Gernot Meier
- 6 July 2023
- Published by Oekom Publishers GmbH in TATuP - Zeitschrift für Technikfolgenabschätzung in Theorie und Praxis
- Vol. 32 (2) , 58-61
Open AccessDigitalisierung, Theologie und Technikfolgenabschätzung: Interview mit Gernot Meier
- Open Access
- 20 December 2021
- Published by Oekom Publishers GmbH in TATuP - Zeitschrift für Technikfolgenabschätzung in Theorie und Praxis
- Vol. 30 (3) , 30-36
Artificial intelligence in human genomics and biomedicine
The increasing availability of extensive and complex data has made human genomics and its applications in (bio)medicine an at tractive domain for artificial intelligence (AI) in the form of advanced machine learning (ML) methods. These methods are linked not only to the hope of improving diagnosis and drug development. Rather, they may also advance key issues in biomedicine, e. g. understanding how individual differences in the human genome may cause specific traits or diseases. We analyze the increasing convergence of AI and genomics, the emergence of a corresponding innovation system, and how these associative AI methods relate to the need for causal knowledge in biomedical research and development (R&D) and in medical practice. Finally, we look at the opportunities and challenges for clinical practice and the implications for governance issues arising from this convergence.
- 20 December 2021
- Published by Oekom Publishers GmbH in TATuP - Zeitschrift für Technikfolgenabschätzung in Theorie und Praxis
- Vol. 30 (3) , 30-36
Open AccessArtificial intelligence in human genomics and biomedicine
The increasing availability of extensive and complex data has made human genomics and its applications in (bio)medicine an at tractive domain for artificial intelligence (AI) in the form of advanced machine learning (ML) methods. These methods are linked not only to the hope of improving diagnosis and drug development. Rather, they may also advance key issues in biomedicine, e. g. understanding how individual differences in the human genome may cause specific traits or diseases. We analyze the increasing convergence of AI and genomics, the emergence of a corresponding innovation system, and how these associative AI methods relate to the need for causal knowledge in biomedical research and development (R&D) and in medical practice. Finally, we look at the opportunities and challenges for clinical practice and the implications for governance issues arising from this convergence.
- 12 November 2019
- Published by Springer Nature
- p. 67-88
Visionary Practices Shaping Power Constellations
- 12 November 2019
- Published by Springer Nature
- p. 67-88
Visionary Practices Shaping Power Constellations