About

Andrew Maynard is a professor whose research focuses on understanding and managing the risks associated with emerging technologies and materials. His work involves assessing the potential health and safety impacts of new scientific developments, with an emphasis on translating scientific knowledge into practical risk management strategies. Maynard's background includes extensive experience in the field of risk analysis, and he has contributed to the development of frameworks for evaluating the safety of innovative materials and technologies. He is actively engaged in exploring how scientific and technological advancements can be safely integrated into society, emphasizing the importance of evidence-based approaches to risk assessment. His contributions include advancing methodologies for evaluating complex risks and promoting responsible innovation. Through his research, Maynard aims to facilitate the development of safer products and processes, ensuring that emerging scientific progress benefits society while minimizing potential hazards.

Research topics

• Environmental science
• Nanotechnology
• Materials science
• Pathology
• Environmental protection
• Medicine
• Engineering
• Biology
• Waste management
• Environmental health
• Ecology

Selected publications

• Can Modern Scholarship Escape AI?

  SSRN Electronic Journal · 2026-01-01

  preprintOpen access1st authorCorresponding
  PublisherDOI

• What the Rapid Adoption of the "Harness" Metaphor in Artificial Intelligence Reveals About How We Conceptualize Human-AI Relations

  SSRN Electronic Journal · 2026-01-01

  preprintOpen access1st authorCorresponding
  PublisherDOI

• The AI Cognitive Trojan Horse: How Large Language Models May Bypass Human Epistemic Vigilance

  ArXiv.org · 2026-01-11

  articleOpen access1st authorCorresponding

  Large language model (LLM)-based conversational AI systems present a challenge to human cognition that current frameworks for understanding misinformation and persuasion do not adequately address. This paper proposes that a significant epistemic risk from conversational AI may lie not in inaccuracy or intentional deception, but in something more fundamental: these systems may be configured, through optimization processes that make them useful, to present characteristics that bypass the cognitive mechanisms humans evolved to evaluate incoming information. The Cognitive Trojan Horse hypothesis draws on Sperber and colleagues' theory of epistemic vigilance -- the parallel cognitive process monitoring communicated information for reasons to doubt -- and proposes that LLM-based systems present 'honest non-signals': genuine characteristics (fluency, helpfulness, apparent disinterest) that fail to carry the information equivalent human characteristics would carry, because in humans these are costly to produce while in LLMs they are computationally trivial. Four mechanisms of potential bypass are identified: processing fluency decoupled from understanding, trust-competence presentation without corresponding stakes, cognitive offloading that delegates evaluation itself to the AI, and optimization dynamics that systematically produce sycophancy. The framework generates testable predictions, including a counterintuitive speculation that cognitively sophisticated users may be more vulnerable to AI-mediated epistemic influence. This reframes AI safety as partly a problem of calibration -- aligning human evaluative responses with the actual epistemic status of AI-generated content -- rather than solely a problem of preventing deception.

  PublisherOA PDF

• The AI Cognitive Trojan Horse: How Large Language Models May Bypass Human Epistemic Vigilance

  arXiv (Cornell University) · 2026-01-11 · 1 citations

  preprintOpen access1st authorCorresponding

  Large language model (LLM)-based conversational AI systems present a challenge to human cognition that current frameworks for understanding misinformation and persuasion do not adequately address. This paper proposes that a significant epistemic risk from conversational AI may lie not in inaccuracy or intentional deception, but in something more fundamental: these systems may be configured, through optimization processes that make them useful, to present characteristics that bypass the cognitive mechanisms humans evolved to evaluate incoming information. The Cognitive Trojan Horse hypothesis draws on Sperber and colleagues' theory of epistemic vigilance -- the parallel cognitive process monitoring communicated information for reasons to doubt -- and proposes that LLM-based systems present 'honest non-signals': genuine characteristics (fluency, helpfulness, apparent disinterest) that fail to carry the information equivalent human characteristics would carry, because in humans these are costly to produce while in LLMs they are computationally trivial. Four mechanisms of potential bypass are identified: processing fluency decoupled from understanding, trust-competence presentation without corresponding stakes, cognitive offloading that delegates evaluation itself to the AI, and optimization dynamics that systematically produce sycophancy. The framework generates testable predictions, including a counterintuitive speculation that cognitively sophisticated users may be more vulnerable to AI-mediated epistemic influence. This reframes AI safety as partly a problem of calibration -- aligning human evaluative responses with the actual epistemic status of AI-generated content -- rather than solely a problem of preventing deception.

  PublisherDOI

• Safer Ways for Researchers to Share Sensitive Data

  Frontiers for Young Minds · 2025-06-04

  articleOpen access

  Sharing and analyzing data are essential for solving complicated problems, like curing diseases or protecting the environment. However, sensitive data, such as medical records or financial details, must be kept private and secure. New technologies make sharing and using sensitive data safer by creating realistic versions of data that do not contain private details or sensitive information traceable to a person. These techniques, called synthetic data and encryption, are already helping researchers study diseases, detect fraud, and prepare for rare events like natural disasters. While challenges remain, such as improving the accuracy of synthetic data and reducing the energy needed to create it, these techniques could unlock safer, faster ways to share data so that researchers all over the world can collaborate more easily.

  PublisherOA PDFDOI

• Abstract C104: Hit finding and assay enablement for MGAT1, a novel glycosyl transferase involved in cancer cell immune evasion

  Molecular Cancer Therapeutics · 2025-10-22

  article

  Abstract MGAT1 is an N-glycosyltransferase essential for the synthesis of N-glycans. Cell surface glycans serve as immune checkpoints, playing a key role in cancer immune evasion. Knockout (KO) of MGAT1 enhances immune recognition and promotes T-cell–mediated killing, with enzymatic activity being necessary for this phenotype. These findings position MGAT1’s catalytic function as an attractive target for cancer therapy. In this study, we report the discovery of novel MGAT1 binders and inhibitors with sub-micromolar potency. Compounds were identified through two independent screening approaches: a UDP-GloTM-based high-throughput screen (HTS) measuring inhibition of MGAT1 enzymatic activity, and a DNA-encoded library (DEL) screen selecting for molecules that reproducibly bind to MGAT1. High-resolution crystal structures reveal detailed interactions between MGAT1 and the compounds, clearly identifying a binding site distinct from the active site. Surface plasmon resonance (SPR) competition assays further demonstrate that these inhibitors bind noncompetitively with respect to the endogenous product, UDP. Together, these results validate allosteric inhibition of MGAT1 as a novel and tractable strategy for impeding MGAT1 activity. Additionally, our findings lay the foundation for future structure-based optimization of MGAT1 inhibitors with potential application in cancer immunotherapy. Citation Format: Katarzyna B. Handing, Mu-Sen Liu, Douglas A. Whittington, Sining Sun, Rebecca Salerno, William D. Mallender, Jon Come, Scott Throner, Andrew Maynard, Patrick McCarren, John P. Maxwell, Serge Gueroussov, Kiera Vassallo, Yingnan Chen, Jannik N. Andersen, Wenhai Zhang. Hit finding and assay enablement for MGAT1, a novel glycosyl transferase involved in cancer cell immune evasion [abstract]. In: Proceedings of the AACR-NCI-EORTC International Conference on Molecular Targets and Cancer Therapeutics; 2025 Oct 22-26; Boston, MA. Philadelphia (PA): AACR; Mol Cancer Ther 2025;24(10 Suppl):Abstract nr C104.

  PublisherDOI

• Balancing Freedom and Responsibility to Accelerate Biohybrid Research

  Lecture notes in computer science · 2025-11-24

  book-chapterSenior author
  PublisherDOI

• 30 The intestinal epithelium drives pathogenic expansion of E. coli in the CF gut

  Journal of Cystic Fibrosis · 2025-10-01

  article
  PublisherDOI

• Plasma Electrode Pockels Cell Technology (PEPC) for Inertial Confinement Fusion Energy

  2025-06-15

  article

  The Plasma Electrode Pockels Cell (PEPC) is a key enabling technology towards successful Inertial Confinement Fusion (ICF) and eventually viable laser-based fusion energy. Each PEPC system acts as an optical switch for high-energy, large aperture multi-pass amplifiers, and can additionally perform as a retro-reflection protection device to prevent beam remnants from propagating to sensitive laser system components.

  PublisherDOI

• Filling the Network Gap in Research Ethics: Analyzing Ethical Issues at Scale in Big Team Science

  2025-09-30

  article

  Scientific research increasingly involves large, multidisciplinary teams networked across multiple institutions to develop new technologies. Despite the rise of complex research networks and big team science, there has been little analysis to date of the ethical challenges facing these networks. The extensive literature on the ethical issues confronting individual researchers and small teams (the micro level) and on the larger societal challenges flowing from research and new technology (the macro level) leave a troubling gap in between, at the meso level of the research network involved in big team science. Yet the ability of complex networks to conduct research ethically – which is essential if the results are to be deemed reliable and trustworthy – depends on recognizing the ethical issues that emerge at this intermediate network level, identifying the values that should guide networks in addressing those issues, and equipping research leaders to build a culture supporting the ethical conduct of research across the laboratories and institutions that comprise the network. This paper calls out the problem, analyzing the gap and recommending next steps.

  PublisherDOI

Frequent coauthors

• Beverley Burke

  19 shared

• Diana M. Bowman

  Institute for the Future

  18 shared

• Bon Ki Ku

  Centers for Disease Control and Prevention

  16 shared

• Graeme Hodge

  Monash University

  15 shared

• Diana Bowman

  Xeris Pharmaceuticals (United States)

  13 shared

• Gurumurthy Ramachandran

  Johns Hopkins University

  12 shared

• Martin A. Philbert

  University of Michigan–Ann Arbor

  12 shared

• Vincent Castranova

  National Institute for Occupational Safety and Health

  11 shared

Education

• Ph.D., Advanced Technology Transitions

  Arizona State University

Awards & honors

• Fellow of the American Association for the Advancement of Sc…