Scientific discovery: a new science
How does a discovery come about? For decades this question has been studied as a science of its own: a theory of epistemic systems, set out in Die Kunst wissenschaftlicher Entdeckungen (1995), and the reconstruction of real discoveries from their documents. Above all Hans Krebs' and Kurt Henseleit's discovery of the urea cycle, reconstructed from their laboratory notebooks with computer models together with Michael May and edited and analysed with Kärin Nickelsen; and Kepler's Astronomia Nova.
Famulus turns this science into a working environment. Causal reasoning reads from the cases which factors are causally relevant; hypotheses are proposed, tested against the cases and revised; colleagues work on the same material; every sentence of a publication is checked against its evidence. Each step keeps its grounds, so a discovery can be followed, repeated and taken further.
The studies behind it are listed under Further reading on this page.
Computational Epistemology: Foundations Kepleriana
Causal reasoning: what is causally relevant to what
96 pages in three volumes
A short circuit is causally relevant to a fire although it is neither sufficient nor necessary for it; ice-cream sales rise with drownings and are not causally relevant to them. One definition, going back to Graßhoff and May (2001), separates the two: a factor is causally relevant to an outcome when it is an indispensable part of a cluster of conditions in a minimal theory of the outcome, however far the record of factors is widened. No counterfactuals, no interventions, no probabilities.
Three volumes develop it, from qualitative inference through causal hypotheses from experimental data to the casebook of the tradition. The causal visual language CVL draws every minimal theory as a diagram that is its formula, and one proof calculus writes all regularity theories, from Mackie to the present, as a single family of inference apparatuses.
Volume I Volume II Volume III CVL Inference Apparatus
Hypotheses that learn from their failures
The Intuition Laboratory turns the description of a situation into factors and a table of cases, derives the causal verdict and sets it against the intuition. Where the two clash, a learning loop revises the causal model with a fixed set of epistemic operators: widening the set of factors, marking factors, cutting back to the typical case.
The Logic of the Casebook shows which intuitions a theory can recover, which need the course of a single episode, and where any theory built on structure and instantiation alone must stop. In the probe lab, the proposition of every stored sentence is set against a probe proposition: does it entail it, contradict it, stand independent of it, or explain it by abduction?
The Intuition Laboratory The Logic of the Casebook
Working together on the same material
Famulus Notebooks are shared documents: questions to a collection, their answers and their checks stand side by side and can be run again by anyone in the project. Projects have members with roles; a public project can be read without an account, and readers can leave annotations in the margin.
Notebooks can draw on collections and computations on other machines of a project, so a research group works on one shared stock of sources. The Kepler notebooks of Kepleriana show it at work: Kepler's lex naturae and his way from the anima motrix to the vis motrix.
Famulus Notebook Kepleriana
Publishing, every sentence with its grounds
The Famulus Lektor reads a manuscript sentence by sentence. A sentence is factually validated when the propositions it expresses follow from evidence, none is refuted and it contradicts nothing said earlier; the author sees for every sentence the grounds that carry it.
Colleagues are invited through a shared link and comment directly on the paragraphs; the author answers in the same place and revises the text with its evidence in view. Papers appear together with their sources, every citation an address that leads to the passage.
Famulus Lektor Sachliche Validität
An answer is the conclusion of an argument
Famulus answers a question to a collection with the conclusion of an argument: addressed facts as premises, declared rules of inference, and critical questions that could defeat the conclusion. Each premise carries its fr: address, so it can be checked one by one.
Validation is a justification status, not a truth predicate. It says that the conclusion follows from premises of a declared strength. The validity tree shows, for every answer, which conclusions lose their justification when a premise is corrected.
Begründete Antworten Validitätsbaum
Computational epistemology: evidence and its reach
21 pages, now public
Evidence and justification are structured differently in the sciences of natural phenomena and in the humanities, yet both rest on one apparatus of validity: reconstructing the argument, identifying its premises, and weighing the evidence that supports them. In the natural sciences that means witnessing and causal regularity; in the humanities, source-grounded interpretation and the reconstruction of intentional action.
The paper explains why answering by retrieved text alone is a category error against this background, and how answers with addressed premises, declared rules and searched defeaters avoid it. It also asks how scholars and AI platforms should divide epistemic labour, and how gaps in the evidence become research agendas.
Computational Epistemology: Foundations
Causality told in cases
8 cases with animated diagrams
The regularity theory is also told through cases with animated diagrams, in German and English: the Manchester hooters, ice cream and drowning, the short circuit and the fire, and cases with experimental data. Each narrative works through one case with diagrams and tables.
The Causal Regularity Bundle describes the form in which the theory is published: papers, casebook, visual language, a skill file and an executable engine, read by human scholars as a set of papers and taken in by AI agents as an operational definition of causal inference.
Kausalität The Causal Regularity Bundle
Application: the causal boundaries of Bell's inequality
22 pages, now public
Graßhoff, Portmann and Wüthrich (2005) derive a Bell-type inequality from nine explicit assumptions, with a separate common cause for each correlation. Read through that derivation, Laidlaw's field-ontology model (2026) satisfies the locality assumptions LOC2 and LOC3 on the wing that is measured first, but not on the wing measured later: there the remote setting enters through the entanglement bond, carried by local interactions between distributed fields rather than by a signal between points.
Since a failed route of proof is no proof of non-derivability, the paper states its claim as a proposition with an explicit countermodel condition. It then fixes the empirical frontier from the primary sources and proposes tests of Laidlaw's prediction that survive correlated noise.
Causal Boundaries of Bell's Inequality
All causal papers in one place
8 causal papers, all public
The papers panel on this page lists every causal paper with its PDF: the three volumes, the visual language and the bundle, and the three papers released with this issue, the synthesis of the casebook, the Intuition Laboratory and the model definitions. The guide sorts them for researchers and for method builders.
Guide
Famulus Science by mail
Famulus Science can be subscribed to on its own or together with the other members of the family. Enter your address in the box at the foot of this page; a mail asks you to confirm, and only then are you on the list.
Each mail carries only the link to the new issue in your language, with one link per news item, and a link to unsubscribe.
