What Would It Mean for Machines to Do Research?

A Conceptual Exploration Drawing on Accounts of Human Inquiry

Shiro Takagi

This is the full text of the book in HTML, one page per chapter. It is also available as a PDF (English, Japanese) and on Zenodo (doi.org/10.5281/zenodo.23851153).

Abstract

What would it mean for machines to do research, and what might change in research when machines do it? Reports of results produced by machines now arrive in quick succession, and with them disputes over what exactly was solved, to whom a result is new, and which part of the work the machine did. This book is a conceptual exploration of these questions. It starts from a thin working characterization: relative to a society S, research is an attempt to bring about epistemic achievements new to S. This lets the one doing research be a machine and lets S be human, mixed, or artificial. The book then draws on accounts of knowledge, understanding, inquiry, and the social organization of research that were developed with human researchers in mind, and follows, account by account, what each implies when the researcher is a machine. It asks when a machine's output becomes knowledge or understanding, and for whom; how inquiry proceeds before there is an answer, from making a question researchable to locating a failure and revising standards; when a configuration can be credited with conducting research and with autonomy, and whose knowledge results; and whether the allocation of inquiry, criticism and certification, and authority over decisions can move inside an artificial configuration. Two things come into view. The claim that a machine is doing research divides into claims that can hold or fail separately. And much that was settled without being decided in human research, because researchers were persons divided among many, becomes something that can be decided: set by builders, shaped by the system itself, or lost without notice. Who holds a belief, how independent critics are, who answers for an error, and against what a revision of standards is measured are examples. The book does not settle these questions; it tries to name what research has rested on, so that they can be asked.

• The text of this book was written mainly by the AI agents Claude (Anthropic) and Codex (OpenAI), working under the author's direction.
• Historical and contemporary cases are used to bring the argument into contact with research practice, not as a representative sample of all research practice.
• The book does not claim that any single definition of research supplies an objective to be optimized in building artificial systems that do research.

Contents

• Introduction
  • 1. A list of operations conceals a conceptual choice
  • 2. Existing labels do not identify one target
  • 3. Why disagreement does not remove the need for characterization
  • 4. A working characterization can be purposeful without being arbitrary
  • 5. Why this book begins from research rather than science alone
  • 6. The route of the argument
• Part I — What Is Research Taken to Be?
• Chapter 1 — A Working Characterization of Research
  • What the characterization needs—and what it does not
  • 1. A thin orientation toward epistemic achievement
  • 2. What the terms are intended to do
  • 3. Attempt separates research activity from its success
  • 4. Bringing about an achievement is neither making an output nor crossing a single threshold
  • 5. Epistemic achievement requires making clear what was achieved
  • 6. New to S makes the unstated reference explicit
  • 7. What follows comparatively directly from the working characterization: being directed at something unsettled
  • 8. What does not follow
  • 9. The debts the characterization leaves open
• Part II — What Would Count as Epistemic Achievement?
• Chapter 2 — When Does a Result Become Knowledge?
  • Epistemic Change, Achievement, and Knowledge
  • 1. From a novel result to an epistemic achievement
  • 2. Process states: what sort of epistemic attitude is at issue?
  • 3. Truth and fallibility
  • 4. What makes a true success epistemic rather than accidental?
  • 5. What verification contributes
  • 6. What is social about knowledge: the reference for novelty, production, possession, and retention
  • 7. The unread artificial discovery
  • 8. Epistemic change before and around knowledge
  • 9. What the epistemological accounts require of artificial research
• Chapter 3 — When Is Understanding Achieved, and by Whom?
  • Knowledge, Explanation, and Opaque Artificial Inquiry
  • 1. Knowledge and understanding as achievements that research may seek
  • 2. Prediction and explanation are related but non-identical achievements
  • 3. What would count as understanding?
  • 4. Knowledge without explanatory understanding
  • 5. Opacity has an address
  • 6. Who understands?
  • 7. What understanding is valued for and what it may enable across time
  • 8. Why a local achievement does not settle continuing inquiry capacity
  • 9. What accounts of understanding require of artificial research
• Part III — How Does Unfinished Inquiry Move?
• Chapter 4 — What Does It Mean for a Machine to Ask?
  • 1. Dividing the claim that "a machine asks"
  • 2. What is asked, and what counts as an answer
  • 3. Understanding a question and making a question tractable
  • 4. How are questions formed and recast within inquiry?
  • 5. The presuppositions of questions, and how questions close
  • 6. Who asks?
  • 7. What is unknown to A and what is unknown to S
  • 8. What does the claim that "a machine generated questions" show?
  • 9. Viewing research from questions, and how this book views it
• Chapter 5 — How Possibilities Become Available
  • 1. Descriptions of procedure and the work they do not show
  • 2. No single method, no return to mystery
  • 3. Making candidates and representations available
  • 4. How candidate formation, pursuit, and appraisal are related
  • 5. Heterogeneous operations, comparable functions
  • 6. Material, embodied, temporal, and distributed inquiry
  • 7. Changes in the possibilities of inquiry
• Chapter 6 — What Does a Result Bear On? Testing, Failure, and Revision
  • 1. From a result to what it bears on
  • 2. When a result runs against expectation
  • 3. When a check is passed
  • 4. Diagnosis across a sequence of results
  • 5. When the standards themselves are in question
  • 6. What the process account passes forward
• Part IV — What Could Make an Artificial Configuration a Researcher, and for Which Society?
• Chapter 7 — When Does an Artificial System Count as a Researcher?
  • Attributing Contribution, Conduct, Capacity, and Status, and Comparing Autonomy
  • 1. The attribution question
  • 2. Conduct: where the organization of inquiry resides
  • 3. Capacity, status, and credit
  • 4. Autonomy as an indexed comparison
  • 5. From attribution to construction and evaluation
• Chapter 8 — Whose Knowledge?
  • Human, Mixed, and Artificial Epistemic Societies
  • 1. Actor, knowledge bearer, and reference society
  • 2. How the world becomes available to an inquirer
  • 3. Nonhuman research and epistemic alignment
  • 4. Two horizons and what Part IV leaves
• Part V — Which Relations around Research Can Move inside an Artificial Configuration?
• Chapter 9 — What Relations Surround Research for S?
  • Roles, Value, Evaluation, and Control
  • 1. Where the attribution of research meets the research system
  • 2. Three relations and the parties who hold them
  • 3. Valuable for S
  • 4. What evaluating involves
  • 5. Control, and the powers that limit it
  • 6. How Part V proceeds
• Chapter 10 — How Is Inquiry Allocated?
  • Division of Labor, Diversity, and Whose Unknowns
  • 1. Why allocation arises, and what makes it hard
  • 2. The philosopher-monarch's two powers
  • 3. Steering without motives: signals
  • 4. Which lines are kept alive, and for how long?
  • 5. What human allocation supplied without deciding it
  • 6. Whose unknowns?
  • 7. What Chapter 10 has found
• Chapter 11 — Criticism and Certification for S
  • Defeat, Independence, Trust, and Correction
  • 1. Which evaluation this chapter examines, and why
  • 2. Independence on the error-detection view
  • 3. Where difference comes from when critics can be made
  • 4. Which errors matter: values in certification
  • 5. Reliance, testimony, and trust in certification
  • 6. When S becomes a layperson
  • 7. What changes in criticism: cost, range, judgment, and arrangement
  • 8. Criticism over time: proxies and standards, evaluating the new, and correction
  • 9. Can criticism for S be organized inside an artificial configuration?
• Chapter 12 — Who May Decide, and Who May Revise?
  • Standing, Revisability, and Relations to Parties Outside
  • 1. Can authority be located inside an artificial configuration?
  • 2. Decisions and decision rights
  • 3. Grounds of authority, and where the epistemic and the normative meet
  • 4. Standing that is more than nominal
  • 5. When A and S are both artificial
  • 6. Keeping authority revisable
  • 7. Decisions that remove the capacity to revise
  • 8. What Chapter 12 has found
  • 9. What Part V has found
• Chapter 13 — Conclusion: What It Means for a Machine to Do Research
  • Research, Artificial Actors, and the Organization of Inquiry
  • 1. Four questions
  • 2. Taking research as the working characterization does
  • 3. When, and in whom, are knowledge and understanding realized?
  • 4. What goes on in research, an undertaking that proceeds before there is an answer?
  • 5. Who is doing the research, and whose do its achievements become?
  • 6. Reorganizing the relations of research
  • 7. Naming what was settled without being decided
• References

Citation

Shiro Takagi. 2026. What Would It Mean for Machines to Do Research? A Conceptual Exploration Drawing on Accounts of Human Inquiry. Zenodo. https://doi.org/10.5281/zenodo.23851153

© 2026 Shiro Takagi. Licensed under the Creative Commons Attribution 4.0 International License (CC BY 4.0), https://creativecommons.org/licenses/by/4.0/. Quoted material from other works is excluded from this license and remains under the rights of its holders.