Max Planck Institute for Innovation and Competition

Abstract Epigenetic clocks and scores have been investigated as potential biomarkers of later-life health outcomes following early-life exposures. In pediatric settings, DNA methylation (DNAm) is often measured in saliva; however, most clocks and scores have been trained in adult blood. Therefore, we assessed the performance, correlation, longitudinal stability, and association with early-life adversity (ELA) for established epigenetic measures in matched pediatric blood and saliva samples. Leveraging the Kids2Health cohort of 291 children (3-12 years, 56% exposed to ELA), we assessed DNAm (Illumina EPICv2) from matched blood and saliva. We calculated 22 commonly used epigenetic measures (chronological and biological clocks, scores for body mass index [BMI], C-reactive protein [CRP], cognition, maternal smoking, telomere length) and compared them with corresponding measured phenotypes (chronological age, BMI, CRP, IQ, telomere length). Overall, performance of epigenetic clocks and scores in children varied widely. Nine epigenetic measures were significantly and equally correlated with their corresponding phenotype across tissues. Six measures were highly correlated between blood and saliva (r≥0.7). All age acceleration estimates showed low to moderate cross-tissue correlations (r=0.20–0.68). Epigenetic scores indicating lower cognitive ability and elevated inflammation were associated with ELA and low SES in both tissues. We additionally provide epigenome-wide blood-saliva correlations across 815,069 CpGs. The results indicate limited generalizability of adult-trained epigenetic clocks and scores to pediatric blood and saliva, even when accounting for cell type composition. We advise caution for cross-tissue and cross-age-range applications of epigenetic measures in research and clinical settings and provide a resource to optimize epigenetic biomarkers in children.

Our study examines how individual characteristics—economic versus achievement-based motives, gender, and experience—moderate the “performance revision effect” in tournament-based crowdsourcing competitions. This effect refers to a phenomenon in which contestants reduce their effort when competing against significantly higher-ability opponents. Using data from Topcoder, a leading crowdsourcing platform, we conducted a quasiexperimental study with 1,677 coders in 38 single-round matches. Our regression discontinuity design exploits Topcoder’s skill-based divisions to assess contestants’ responses to differing opponent abilities. The results confirm the performance revision effect, revealing an average performance decline of 20% when contestants face higher-ability opponents. Moreover, female and more experienced participants show a stronger response to the performance revision effect than their male and less-experienced peers. Our findings contribute to the crowdsourcing literature by highlighting the boundary conditions of the performance revision effect and by quantifying the performance implications of contest designs for different contestants, allowing platform operators to make data-driven cost-benefit decisions about contest design to mitigate performance losses. Supplemental Material: The online appendix is available at https://doi.org/10.1287/stsc.2023.0068 .

Abstract This article addresses the legal tensions between artificial intelligence (AI) development and copyright law, exploring policymaking on the use of copyrighted data for AI training at the input level and the generation of AI content at the output level. Currently, global policy responses have heavily focused on the input level—whether AI can lawfully incorporate copyrighted data for AI training purposes. Jurisdictions such as the EU, the UK, the US, China and Japan adopt varied approaches. By comparing and examining different policy versions, this article proposes shifting the focus from input restrictions to output regulation, a policy strategy referred to as ‘input out, output in’. It suggests that AI training should generally be lawful, while regulatory guardrails should apply to outputs that may compete directly with copyrighted works and deprive rightsholders of their deserved revenues. To harmonize the relationship between copyright holders and AI developers, key policy tools may include promoting transformative use, proper quotation and attribution, a Creative Commons-style framework and the safe harbour mechanism. This output-focused approach seeks to create positive-sum outcomes for copyright holders, AI developers and public information consumers. By ensuring free access to training data while moderating AI-generated content, the proposal supports innovation, protects creators’ interests and enhances public access to quality information, ultimately promoting a balanced and sustainable legal framework.

This contribution critically examines the absence of a comprehensive transnational competition law regime and explores the role of competition authority cooperation as an alternative form of governance. It examines the political economy of past harmonisation efforts, current transnational approaches, and the viability of a global competition law framework. Grounded in the public autonomy–private rights paradigm, the authors assess the tensions between technocratic governance and democratic legitimacy, which extend from domestic competition policies to international cooperation mechanisms such as the International Competition Network. The authors explore the potential of a Plurilateral Agreement on Competition to mitigate cross-border externalities and enhance global welfare and critically evaluate whether the efficiency gains such an agreement might yield justify the erosion of public autonomy inherent in technocratic structures. Highlighting the importance of regional diversity in competition law frameworks, the authors contend that context-sensitive, institutionally pluralistic approaches provide a more effective means of balancing public autonomy with the protection of private rights while adapting to individual jurisdictions’ unique economic and political contexts.

Abstract Carbon credits from avoided deforestation projects have been criticised for overcrediting. Here, we investigate how the type of organisation implementing and validating the projects affects outcomes. We show that non-profits are more than twice as likely as for-profits to deliver carbon credits that represent real emission reductions.

This analysis shows that the decision of the German Federal Court of Justice in the DABUS case provides for a surface-level approach to the much-debated issue of inventorship in the context of artificial intelligence (AI) applications. While the Court accepted the additional information concerning the invention’s genesis on the inventor designation form as being in conformity with the existing (procedural) law, it disregarded the substance of this information, which prima facie raises doubts as to whether the designated natural person is indeed the inventor, particularly given the highly contentious circumstances of Thaler’s case. Although the Court upheld the merit-based notion of inventorship and the sufficient contribution requirement, its formalistic treatment of the inventor designation – such that “any human would do” – risks reducing the inventor designation to a legal fiction, allowing trivial human involvement to qualify as inventorship, especially in cases where unjustifiable claims are unlikely to be contested. The pending legal question left unanswered by the Court is: Under what conditions can, and should, a human using AI in the inventive process fulfil the requirement of sufficient contribution to merit the title of inventor, and how should the rules on inventor designation be applied to safeguard their ratio legis? The decision can be read as encouraging innovators to apply AI in technical problem-solving, leading to potentially patentable inventions, as it ensures that the contribution of AI to finding a technical teaching – regardless of how substantial it might be – would not preclude granting a patent to a natural person. This suggests that the Court prioritised the potential patentability of AI-assisted inventions over the merit-based justification for inventorship.