Now Show the Grounds

A text lies on the table again. Beside it, a model that we can read, question, and change. In the prologue we did not yet know which differences between them would matter. Now we can trace those differences in the model.

It all started with water, which can give a proton to another particle and accept one. The textbook calls particles that play both roles amphiprotic. We identified the subject, recorded two pieces of information, and watched a rule link them to the conclusion.

Then one piece of information went missing. The conclusion disappeared — but that did not establish the opposite. Conflicting messages, exceptions, dates, sources, and readings entered the picture. The question “what does the answer rest on?” grew sharper.

Now it may sound like this: Who reported this information? Does it concern the same object? Which fragment supports the rule? Does the rule apply as of the chosen date? Which definition are we using? What happens if one of the premises is removed?

These questions now have places in the model: assertion, rule, source reference, context field, explanation branch. Not every path settles the matter. But each gives us a place to start checking a claim or discussing a disagreement.

What remains our responsibility

The model does not relieve us of responsibility for the interpretation. Someone picked a condition out of the text, chose a vocabulary, and judged the information sufficient for the conclusion. Someone judged the formula applicable to the experiment, or accepted a message as a trustworthy ground.

Some of these decisions can be written down explicitly. Others hide in relation names, in the choice of subject, in what the model never asks. So working with grounds does not end when a file is accepted or every test passes.

In the amphiproticity model, we had to explain why the two abilities are checked jointly. In the work calculation, we had to name the conditions the formula needs. In interpreting a document, we had to distinguish a chosen reading from an established claim about the text’s history.

An explicit model puts these decisions on the table. Anyone can point at a line, propose another, and compare the consequences. But noticing a missing assumption, checking the source, and judging the applicability limits remains human work.

A ground to return to

Talking about the grounds of knowledge offers no promise of an infallible answer. It starts with a readiness to tell apart what we observed, what we accepted, what we derived, and what we have not yet sorted out.

The model helps keep these differences when it is handed over. The recipient sees not only the outcome but the path to it: information, choices, rules, constraints. They can repeat the calculation, dispute a premise, and check what changes after replacing it.

Repeatability does not make a statement true. It gives further discussion a shared basis: we make sure we are discussing the same model and the same calculation. Then disagreement can narrow — to the source, an assumption, a rule, or the question — instead of dissolving into two incomparable answers.

When you are handed a model, ask to be shown its grounds. When you hand over your own, show its grounds first.

A handover card for the model

Enclose a short card with the set you hand over. It helps the recipient find the essentials before reading every file.

For the “Water” example the question is about the amphiproticity of urn:case:kniga:h2o. The inputs are the supplied assertions about donating and accepting a proton. The expected answer is TRUE_ONLY, COMPUTED; after removing the assertion about acceptance — NEITHER, COMPUTED. The rule and the source come from the pinned chemical package; the previous chapter gives the commands and shows how to produce the result document.[^epilogue-voda]

Give your text a card of its own, perhaps quite short. What matters is that the person across the table can open the set, find the needed files, and continue the reasoning independently.

[^epilogue-voda]: The full handing-over experiment is given in chapter 18. The original “Water” case holds two assertions; removing the voda-2 block gives the stated NEITHER, COMPUTED answer. This checks the given example.