Humans are Missing from AI Coding Agent Research

Recent progress in AI coding agent research has led to rapid improvements in agents' ability to autonomously perform complex software engineering tasks, from editing large codebases to executing long-horizon development workflows. As these systems make strides, however, the primary bottleneck to practical usefulness increasingly shifts away from pure task-solving capability, and toward challenges in how users communicate with, supervise, and trust agents. In this position paper, we argue for a reorientation from autonomous to human-centered coding agents: systems designed not only to complete tasks, but to collaborate effectively with people. We identify four core interaction-level dimensions that characterize the human-agent task-solving loop: task alignment, verifiability, steerability, and adaptability. Finally, we outline concrete research directions to advance these dimensions, including user-involved coding environments, comprehensive verification mechanisms, and principled measures of human-agent interaction quality.

Reinventing Discovery

How the internet and powerful online tools are democratizing and accelerating scientific discovery

Human learning is an understudied but promising lever for boosting human–AI synergy

Humans collaborating with AI hold the promise of achieving superior outcomes compared to either acting alone (i.e., human–AI synergy). However, the conditions that facilitate such synergy when humans are advised by AI are not well understood. A recent meta-analysis showed that, on average, human–AI combinations do not outperform the better individual agent. We argue that this pessimistic conclusion arises from insufficient attention to human learning in experimental designs. To substantiate this claim, we reanalyzed all 74 studies included in the original meta-analysis and found that most previous research overlooked design features that foster human learning (e.g., outcome feedback to participants). Our reanalysis further revealed that studies providing outcome feedback show tentatively higher synergy than those without outcome feedback. Crucially, feedback paired with AI explanations was associated with positive synergy, while explanations without feedback were associated with negative synergy—suggesting that explanations improve synergy mainly when humans can learn to verify the AI’s reliability through feedback. Our reanalysis suggests that the current literature underestimates the potential of human–AI collaboration because it predominantly relies on paradigms that do not facilitate human learning, thus hindering humans from effectively adapting their collaboration strategies. However, experiments directly varying learning opportunities are needed for stronger, causal conclusions. We advocate for a paradigm shift in human–AI interaction research that explicitly addresses human learning and thus enhances our understanding of and support for successful human–AI collaboration.

AI for science needs reasoning, not just data

AI agents that can model the human process of research will accelerate discoveries in science.

MOSAIC — Modular Open Science, AI & Collective Intelligence

MOSAIC is an R&D program to build a new medium for large-scale, AI-augmented networked research.

Reimagining research papers as interactive and reliable AI agents

Here we introduce Paper2Agent, an automated framework that converts research papers into artificial intelligence (AI) agents. Paper2Agent transforms research output from passive artefacts into active systems that accelerate use and discovery. Conventional research papers require readers to understand and adapt the paper’s code, data and methods to their work, creating barriers to dissemination and reuse. Paper2Agent addresses this challenge by converting a paper into an AI agent that functions as a virtual corresponding author, exposing its manuscript, supplementary materials, datasets, code and workflows as active, agent-native knowledge rather than static text. It analyses the paper and codebase using multiple agents to construct a model context protocol (MCP) server, then generates and runs tests to refine and increase robustness of the MCP. These paper MCPs can be connected to a chat agent (such as Claude Code) to carry out complex scientific queries through natural language while invoking tools and workflows from the paper. We demonstrate Paper2Agent’s effectiveness through case studies. Paper2Agent created an agent that leveraged AlphaGenome1 to interpret genomic variants and agents based on Scanpy2 and TISSUE (transcript imputation with spatial single-cell uncertainty estimation)3 to conduct single-cell and spatial transcriptomics analyses. We validate that these agents reproduce the results of the original papers and carry out novel user queries. Paper2Agent created multiple agents that collaborate to prioritize a causal gene for psoriasis. By turning static papers into interactive AI agents, Paper2Agent introduces a paradigm for knowledge dissemination and a collaborative ecosystem of AI co-scientists.

Where does the rigor go? Research software and the future of trustworthy science.

Generative AI now makes it dramatically easier to produce something that looks like research: analysis code, figures, literature reviews, even whole pap…

eLife Claim Trees — eLife Claim Trees

Panel-level claim graphs for reproducibility — eLife Claim Trees

Grand Challenges for the Convergence of Computational and Citizen Science Research Workshop Report

This report is an outcome of a Computing Community Consortium (CCC) visioning workshop on Grand Challenges for the Convergence of Computational and Citizen Science Research conducted on April 8-9, 2025, in Washington, D.C. as well as through several precursor virtual input-gathering sessions. These events brought together experts across relevant disciplines to develop a research agenda that brings to fruition the above vision on how humans and machines may team up to solve some of the world's most pressing scientific problems. Citizen science delivers measurable economic and national value. Public participation in scientific research generates millions of dollars in volunteer labor value, extends government agency capacity, and directly supports federal priorities in areas such as disaster management, public health, water, energy, workforce development, and many more. At the same time, 21st-century scientific infrastructure requirements for citizen science (from hardware and cyberinfrastructure to data and computational frameworks) mirror those for computational science more generally. The distributed, collaborative, long-term, and contextual nature of citizen science makes it a demanding real-world use case for a novel robust research infrastructure that accounts for security, privacy, resource adaptability, and transparency. In this report, we outline the key findings, future research directions, and recommendations that emerged from the April 2025 CCC Grand Challenges for the Convergence of Computational and Citizen Science Research Workshop.

AI as Normal Technology

The normal technology frame is about the relationship between technology and society. It rejects technological determinism, especially the notion of AI itself as an agent in determining its future. It is guided by lessons from past technological revolutions, such as the slow and uncertain nature of technology adoption and diffusion. It also emphasizes continuity between the past and the future trajectory of AI in terms of societal impact and the role of institutions in shaping this trajectory.

Is the scientific paper still a fraud?

How we write scientific papers does not reflect how we do science. Their formal structure infers a pre-ordained linear process rather than reflecting the messy creativity of research. This matters in the AI age because it masks the human in the process.

The Robyn Dawes Institute for the Improvement of Science

Making research quality visible so everyone can act on reliable evidence.

The Least Agentic People Alive

The arms race of deferring agency and the dark acquiescence to robotic bureaucracy.

How can citizen science reduce psychological distance to science? Insights from three projects in contested environmental contexts

Citizen science (CS) is a participatory mode of knowledge production, enabling non-scientific actors to contribute to and sometimes contest scientific agendas and interpretations, making it a way to bridge science and society. This paper examines how that potential unfolds by analysing the individual perspectives of citizen scientists through the lens of Psychological Distance to Science (PSYDISC). Drawing on three case studies of contested environmental CS, we identify which contextual aspects of CS shape citizen scientists' experiences of social, spatial, temporal, or hypothetical distance to relevant science, and how these experiences may relate to trust. Our findings underscore the role of science communication as both a channel for dissemination, and as a constitutive element of participatory research; crucial for reducing psychological distance and enabling socially robust knowledge production, especially in contested, policy-relevant science settings.

Nova Scotia’s Experiment in Research That Solves Real Problems

Locally focused research investments enable coalitions of scientists and citizens to serve community needs.

We argue badly, and nothing accumulates. How could we do better? | Reason Commons — Issue Trees & Logical Thinking Process

Millions of people argue every day about the things that matter most — climate change, what to do about AI, how we might build a better world. Some of it is sharp, even insightful. And almost none of it accumulates.

The Ground Truth Institute

Could we engineer scientific revolutions, then spin out radical improvements to everyday life?

Can we measure trust in scientific publications? - LSE Impact

Jonathon Alexis Coates outlines how a constellation of static and dynamic indicators could provide a means for assessing the trustworthiness of published research