Skip to content

What is Research Information Systems Engineering?

Why now: GenAI is becoming research infrastructure

Generative AI is no longer a novelty tool used at the edges of scholarly work. It is becoming part of the infrastructure of scientific work itself — drafting literature reviews, generating analysis code, writing manuscript sections, simulating peer review, proposing hypotheses, and (in a growing class of systems) running parts of the discovery loop end-to-end.

Ad-hoc adoption of these systems creates institutional risks the existing apparatus of scholarship was not designed to absorb: hallucinated citations enter the published record, replication credit drifts to opaque pipelines, reviewer labor reallocates without policy, and the line between "human contribution" and "AI contribution" blurs in ways that current authorship and disclosure norms do not address.

The institutional response — guardrails set by journals, funding bodies, departments, and the broader research community — needs to be grounded in a systematic understanding of these systems, not in enthusiasm or skepticism alone. That understanding has to span at least five dimensions:

  • Architectures. How the pipeline is composed: single-agent, multi-agent, human-in-the-loop, evolutionary search, RAG-grounded. What the orchestration topology is. Which tools are wired in.
  • Human roles. Who decides, who verifies, who is accountable for the output. Where the augmentation–automation slider sits, and how it shifts across stages of the pipeline.
  • Provenance mechanisms. How each claim in the output can be traced back to its source — the data, the prompt, the intermediate artifact, the code commit, the cited paper, the model version.
  • Failure modes. What plausible-sounding wrong outputs look like (hallucinated citations, unfaithful reasoning, scope drift, fabricated quotes), and how they are detected and contained.
  • Evaluation criteria. How the system's outputs are judged — as code, as artifact, as scholarly contribution — and how those criteria differ from the ones used for traditional human-only research.

Research Information Systems Engineering is the framework that does this systematic work.

Working definition

Research Information Systems Engineering (RISE) is a framework for designing, operating, and governing AI-enabled research systems — information systems whose primary purpose is to produce scholarly knowledge with at least partial mediation by AI agents.

A RISE system is distinguished from neighboring artifacts by three commitments:

  1. Knowledge production as output. The system's terminal artifacts are scholarly: papers, datasets, code releases, replication reports — not predictions, decisions, or recommendations as ends in themselves.
  2. Agentic mediation. The pipeline is at least partially executed by AI agents — LLMs with tools, multi-agent orchestrations, or human–AI hybrids — not solely by human researchers using non-agentic software.
  3. Methodological accountability. The system is designed against research-methodological norms (identification, replicability, peer-review readiness, citation grounding), not only against software-engineering norms.

The three commitments are jointly necessary. A system that produces scholarly outputs without agentic mediation is a research workflow, not RISE. An agentic system whose terminal artifact is a chat response is a chatbot, not RISE. An agentic system that drafts papers but does not engage methodological standards (no identification, no citation grounding, no review) is a style engine — see 1 — useful, but not RISE.

The framework framing matters: RISE is not a single artifact, a single project, or a single discipline. It is the set of design principles, evaluation rubrics, vocabularies, and governance norms that let researchers, builders, and institutions reason about AI-enabled research systems coherently — across the design, operation, and oversight phases of their lifecycle.

Why a separate name?

The umbrella terms in use — AI for science, AI scientist, agentic research, autonomous research agents — each foreground one face of the object. AI for science is too broad: it includes AlphaFold-style prediction systems whose outputs are scientific claims about the world rather than scholarly artifacts. AI scientist anthropomorphizes a system as a researcher, which prejudges the sociotechnical question (cf. 2). Agentic research describes the method but underspecifies the output and accountability dimensions.

RISE names the engineering, methodological, and governance framework that sits across these — borrowing from information-systems design science, e-Science research infrastructure, and AI4Science — and treats the agentic research pipeline itself as its primary object of design, operation, and oversight.

What RISE is not

  • Generic AI-for-science tooling (e.g., protein-structure prediction) that produces scientific predictions rather than scholarly artifacts.
  • Pure information-systems theory without a built artifact.
  • AI writing assistants whose unit of output is prose rather than research products.
  • Evaluation infrastructure for AI systems (e.g., aviary) — directly relevant, but not a RISE pipeline.

Scope of this knowledge base

RISE here is treated as a named subfield, not a synonym for "AI for science." The literature surveyed (see Papers) foregrounds the agentic pipeline as the object of design and study. The projects catalog evaluates concrete systems — both owned (E2ER) and external (Sakana AI Scientist, STORM, GPT Researcher, PaperQA2, OpenScholar, MARG, Agent Laboratory, Robin, APE, Reviewer, …) — against a shared rubric.

Sociotechnical commitment

This knowledge base sits, deliberately, in the sociotechnical tradition of information systems research 3. A RISE contribution is not only a technical artifact: it is an intervention in the social system of scholarship. That commitment shapes the evaluation rubric (autonomy level and human-in-loop are first-class), shapes the scope (peer-review tools are first-class projects, not peripheral), and shapes the kinds of papers featured in the literature (the discipline's reception of GenAI is as central as its mechanics).

  • e-Science and research infrastructure. Cyberinfrastructure for large-scale, data-intensive science.
  • AI4Science. Broader — includes scientific prediction (AlphaFold) alongside scholarly-artifact production.
  • Information Systems (design-science tradition). The methodological home of artifact-producing research.
  • Open and reproducible science. Pre-registration, replication, computational reproducibility.
  • Science of science / metascience. Empirical study of how research is produced and disseminated.

See History for the lineages these draw on.

  1. Riemer, K., & Peter, S. (2024). Conceptualizing generative AI as style engines: Application archetypes and implications. International Journal of Information Management, 79, 102824. https://doi.org/10.1016/j.ijinfomgt.2024.102824 ↩

  2. Peter, S., Riemer, K., & West, J. D. (2025). The benefits and dangers of anthropomorphic conversational agents. Proceedings of the National Academy of Sciences, 122(22), e2415898122. https://doi.org/10.1073/pnas.2415898122 ↩

  3. Sarker, S. et al. (2019). The sociotechnical axis of cohesion for the IS discipline: Its historical legacy and its continued relevance. MIS Quarterly, 43(3), 695–719. https://doi.org/10.25300/misq/2019/13747 ↩