B.5.TU - Construct a Working Use of an Unfamiliar Theory

Type: Method-description pattern Status: Draft Normativity: Normative unless marked informative

B.5.TU:1 - Problem frame

Use this pattern when a theory offers a promising way to answer a question, but you cannot yet connect its objects and reasoning to an answer you can use. You may recognize the notation or reproduce a calculation while still being unable to say what to supply, which operation matters, or what the result would let you do.

The work here is to construct one application of the theory. Begin with the question and the available account; obtain and interpret the contribution that question needs. The result can be a construction, prediction, explanation, bound or identified missing operation. A mathematical question about what can be constructed is a possible receiving use, as is an engineering decision.

First useful move. Choose one question the theory might settle. Find a source case or rule that could supply its answer. Say what plays each required part in your case, work the decisive operation, and explain what its result changes. If you cannot make that connection, locate the first missing premise, interpretation or operation.

For example, a theory of composition lets you combine two operations. Does it let both consume one input, or does the combination require two separately supplied inputs? Recovering that distinction can change the proposed implementation before any code is written.

Use an adequate, understood application directly. A specialist or tool may supply a result you can already interpret and use under its conditions. Reconstruct more of the theory when the needed application, criticism, adaptation or explanation remains unresolved.

B.5.TU:2 - Problem

A theory’s descriptions usually distribute its usefulness across definitions, examples, constructions and arguments. A reader may know each familiar word yet be unable to assemble these contributions for a new question. Conversely, a familiar calculation can conceal a change of question: the calculation solves the stated equations, while the receiving work needs a condition those equations omitted.

The difficulty has two connected parts. First, discover what the account makes possible and how to obtain that result. Then apply it to the question that justified the work. An answer can also reveal a new question or a useful limitation of the theory.

B.5.TU:3 - Forces

B.5.TU:4 - Solution

Connect the working question, the theory’s constructive or inferential resources, and the use of the resulting answer. Enter where the unresolved connection begins; reuse contributions already available. The following unfolding can be revisited when its own result changes an earlier choice.

B.5.TU:4.1 - Choose an application worth constructing

State what an answer would help someone understand, obtain, predict, rule out or change. Select one case small enough to work with the available means while retaining the difficulty. For example, “Can these two operations use the same input?” gives composition rules a purpose; “understand category theory” leaves that first use unselected.

Locate a source application serving that question, or the nearest account whose objects and rules could be adapted. State which part is supplied and which you are proposing. A theory can also be explored to discover a useful question: try one of its characteristic operations and inspect what its result makes distinguishable or obtainable.

Determine the contribution you need to learn or obtain. Applying a supported formula can need its inputs and use conditions; changing the formula can need its construction and argument. Explain the needed difference in work before selecting further study.

B.5.TU:4.2 - Recover what the theory lets you work with

For this application, identify the objects, relations, given information, permitted constructions and inferential rules. Read a definition together with an operation that uses it. Ask what can be supplied as an input, what can be produced, and which property or conclusion the rule establishes.

Follow the needed result backward to its prerequisites. Keep shared prerequisites and jointly required inputs visible. When a description names a construction without showing how to perform it, B.5.RC recovers its producing operations. When an inference remains unclear, B.5.RA recovers the reason and transitions needed by the receiving use.

Make an unfamiliar expression operative at the point of difficulty. Determine the arguments of a function, the objects related by a diagram, or the conditions under which a rule applies. A.6.3.RT helps construct a usable expression under the available notation. Understanding a symbol’s name is useful only as far as it helps perform or interpret the operation.

B.5.TU:4.3 - Construct the application

Connect each consequential object and condition to the selected case. Some are supplied by observation or design; some are assumptions of the model; others belong to the theory’s mathematical construction. Keep their roles clear where they affect the conclusion.

Choose the values, boundary conditions, allowed operations and sought output. For a theoretical question, this may be entirely a construction within the theory. For a physical use, explain which system, quantity or interaction the mathematical object represents. C.29 supplies the mathematical correspondence and the comparison of what it preserves, omits or introduces.

Compose the required operations. Check the connection between successive contributions: an operation’s output must supply the next operation’s required input. An integration procedure, for example, needs a function of its integration variable; a formula with unfilled path or parameter arguments has yet to supply that function.

If a source case cannot be adapted by the available rules, identify the particular missing contribution. Another formulation, a specialist result or a developed operation may supply it. A.15.9 supports obtaining and using bounded help; C.39 supports developing a way when none is known.

B.5.TU:4.4 - Obtain and inspect a consequence

Work the selected case far enough to answer the question or expose the gap. Use direct reasoning, a construction, a small computation or an available tool according to the work. Recover the step that makes the consequence follow and the premises it uses.

When computation is involved, interpret its inputs, operations and output. C.29.2 supplies a missing computational formulation; C.29.3 supplies the connection to the system performing it. A successful execution can help locate expression or implementation errors. The argument for what the computation establishes remains tied to the chosen formulation.

Compare with a simple consequence you can inspect when that comparison can reveal an error: a limiting case, a preserved quantity, an input count, a dimensional relation or a small enumeration. Select the relevant comparison; its purpose is to expose a possible failure in this application.

B.5.TU:4.5 - Apply the answer to the working question

Read the answer back in the terms of the intended use. If it gives a construction, determine whether the required inputs and operations are available. If it gives a bound, use the exclusion or allowance that bound supports. If it gives a physical prediction, identify the model and operating conditions under which that prediction concerns the system.

Examine the correspondence at the point where it could change the action. Does a feasible mathematical answer violate a receiving condition? Could two cases treated alike by the formulation require different actions? If so, restore the distinction, change the formulation, or keep a weaker answer that remains useful.

For a new physical reliance, address the model assumptions whose failure could change that use. Reuse adequate knowledge and observations. Choose further checking or acquisition under C.11.DUA when their benefit and burden are the live decision; an unresolved stronger claim need not displace a sufficient conditional answer.

B.5.TU:4.6 - Return the result and choose the continuation

Return the interpreted answer and the conditions its receiver needs. An ordinary explanation, calculation or demonstration can carry it. Preserve a separate account when another participant or later use needs to reconstruct the application.

Stop when the contribution suffices. Otherwise identify what remains: a source explanation, missing operation, changed model, competing theoretical account or newly worthwhile question. B.5.RR follows the effect of a changed premise; B.5 coordinates a changed inquiry.

Use this result to choose the next learning or work contribution. When the work requires one participant to apply or alter the method independently, that capability needs its own performed use under the relevant conditions. A supplied answer can still serve work that does not require that independence.

B.5.TU:5 - Archetypal Grounding

B.5.TU:5.1 - Discover whether one input is enough

A practitioner wants to use two available operations, f: X -> Y and g: X -> Z, on one input. The letters name distinct atomic types. The question is whether the chosen composition rules can produce both results from that input.

In a cartesian account, a copying operation Delta_X: X -> X x X supplies the pair of inputs. First copy, then apply the two operations: (f x g) after Delta_X sends x to (f(x), g(x)). The input of the parallel operation is now supplied.

Compare a resource-sensitive account generated only by f, g, identities, serial composition, tensoring and exchange of factors. Tensoring f with g takes X tensor X to Y tensor Z: it requires two inputs. Each given generator preserves the number of atomic factors, and serial composition, tensoring and exchange preserve that property. These rules therefore cannot construct X -> Y tensor Z. Supplying another input or adding an admissible copying operation changes what can be built. This is a result about the stated generated account.

Now apply the distinction. For a reusable data value interpreted through pure functions, both uses can receive that value. For a physical specimen consumed by an assay, obtain a physical way to supply what each assay needs. Dividing a specimen may supply those inputs if the assays admit the resulting portions. The word “copy” would leave that physical contribution unexplained.

The common method discovers the needed theory operation and returns its consequence to use. The cartesian construction and its contrast with tensor composition come from Baez and Stay, §2.3; the input-count argument makes the stated restricted case inspectable.

B.5.TU:5.2 - Use variational mechanics to answer a motion question

An engineer wants to understand what a variational account predicts for an ideal carriage coasting along a straight horizontal track. Adopt a one-dimensional nonrelativistic free-particle model: positive mass m, negligible resistance and no applied driving force during the interval. Its Lagrangian is L(q, v) = m*v*v/2. For this case, the action is the time integral of that quantity along a proposed path. The physical model supplies this choice of Lagrangian.

The mechanical rule selects paths of stationary action: for every small path variation that leaves the endpoint positions fixed, the first-order change of action must vanish. The calculation below finds that path and shows that, for this free particle, it minimizes the action. Reproducing the calculation uses differentiation and definite integration; the displayed action identity can also be obtained as a supplied mathematical result.

Let the carriage pass q=0 at time 0 and q=d at positive elapsed time T. Recover the operations hidden by “calculate the action”: choose a position function q(t), differentiate it to obtain velocity, evaluate L on those values, then integrate over time. This is the constructive use developed in SICM, §§1.3–1.4. The following one-dimensional case uses it.

Start with q_0(t)=d*t/T, and try paths q_a(t)=d*t/T + a*(t/T)*(1-t/T), where a is a length. Every path has the specified endpoint positions. Its velocity is v_a(t)=(d+a*(1-2*t/T))/T. Substitution and integration give S_a = m*(d*d+a*a/3)/(2*T). For m=1 kg, d=2 m and T=1 s, the straight path has action 2 J s; a=1 m gives 13/6 J s.

This comparison favors the straight path within the chosen family. The general argument is also short. For any continuously differentiable added displacement eta(t) that is zero at both endpoints, the cross term integrates to (m*d/T)*(eta(T)-eta(0))=0. Thus the change in action is (m/2)*integral(eta'(t)^2 dt), which is nonnegative. The straight path minimizes this action among those paths. For a nonzero eta, write this positive change as K. Along the paths q_0+c*eta, the action is S[q_0]+c*c*K, whose derivative at c=1 is 2*K>0. Thus a nonzero displacement from q_0 cannot be stationary. The selected motion has constant velocity d/T.

The application now has an interpreted answer: under the adopted model, passing those endpoint positions in that time entails constant velocity. A computational implementation must evaluate the path derivative in the velocity argument before integrating. It can reproduce the numerical action comparison; the displayed argument supplies the wider conclusion.

Suppose the same two-metre trip must instead begin and end at rest. The constant-velocity answer fails that condition. The useful return is that the undriven free-particle account cannot supply this trip: the design needs acceleration and deceleration, and an account of the forces producing them. The next contribution is that driven-motion model. Increasing the resolution of the same free-particle calculation would retain the missing physical contribution.

B.5.TU:5.3 - Keep a useful bound when the optimizer cannot be enacted

A planner learning optimization must select whole jobs for four available hours. At most one A-job is available, taking three hours for stipulated value 5; at most two B-jobs are available, each taking two hours for value 3. Durations and values add in this constructed problem.

A linear relaxation permits real counts 0 <= x <= 1, 0 <= y <= 2 with 3*x+2*y <= 4. It returns x=1, y=0.5, value 6.5. Recover the argument: the time condition gives y <= (4-3*x)/2, so 5*x+3*y <= 6+0.5*x <= 6.5; the returned pair attains that limit.

For the actual whole-job choice, enumerate the alternatives. With A selected, no B fits and value is 5. Without A, two B-jobs fit and value is 6. Select two B-jobs. The fractional optimizer has still supplied a useful bound: no whole-job arrangement can attain value 7 because every such arrangement is also admitted by the relaxation.

The use of the theory changes with the question. Selecting work needs an attainable arrangement; excluding a target can use a bound. The practitioner needs to recover that relation before deciding whether a further optimization step is useful. B.5:5.4 also works the changed five-hour case.

B.5.TU:6 - Bias-Annotation

The worked cases favor explicit mathematical accounts with small, inspectable operations. Other theories can use qualitative or interpretive inferences; their warranted application conditions and criticism must come from the relevant practice. A convenient numerical example should not decide which kind of explanation the question needs.

Familiar terminology can make a supplied interpretation look independently recovered. Attribute the actual contribution when assessing learning or capability. In ordinary assisted work, use the supplied result at the scope its interpretation supports.

B.5.TU:7 - Conformance Checklist

• The application has a working question and a useful result to obtain or interpret.
• The theory’s objects, operations and premises needed for that application are recoverable.
• The construction connects supplied inputs to required outputs; a missing operation remains a specific gap.
• A source-supported application and a newly proposed adaptation retain their respective grounds.
• The worked consequence states what its construction, argument or computation establishes.
• The receiving use preserves consequential conditions, including any difference between an attainable result and a bound.
• Further study, evidence or specialist work serves a remaining useful contribution at an appropriate cost.
• The return supports an action, understanding, construction or next question; sufficient use can stop.

B.5.TU:8 - Common Anti-Patterns and How to Avoid Them

B.5.TU:9 - Consequences

The practitioner can move from a promising theory to a worked use, or state the contribution that prevents it. The result also makes division of work easier: one participant can supply a model, another a construction or calculation, and another the receiving interpretation.

The cost is recovering enough of the account to assemble and use the application. Reuse supplied explanations and operations where adequate. Detailed derivations, broader source comparison or empirical inquiry are worthwhile when their answers affect the intended work.

B.5.TU:10 - Architectural Rationale

The method joins two directions of reasoning. Working backward from the question reveals the needed contribution. Working forward through the theory tests whether the available objects and operations actually supply it. Returning the consequence to use exposes conditions a correct internal calculation can leave unresolved.

B.5.RC and B.5.RA recover constructions and arguments. Here they contribute to assembling an application whose meaning and needed output may still be unsettled. C.29 constructs and tests mathematical correspondence; it can start before a mathematical object has been selected. This method adds recovery of how the unfamiliar theoretical account becomes usable at that point. A theory can also be used within its own mathematical setting, as in the composition case.

A complete explanation of the theory can be the appropriate learning project. For one urgent use, however, an adequate supplied application or bounded specialist result may settle the question sooner. The choice depends on the work the receiver must perform, including any need to criticize or change the method later.

A small successful application provides something to use and a basis for further inquiry. Broader generalization requires its own argument. If the application exposes a limitation in the theory, the resulting construction or counterexample can become material for developing its objects, rules or questions.

B.5.TU:11 - SoTA-Echoing

How much theory must be reconstructed for a first use? Adapt van Oostrum, Langer and Ay (2025), Introduction and §§1–2, 4: connect the operational question to stated model variables and inference, work an example, and recover deeper derivations when they are needed. The alternative is to use an interpreted supplied implementation. Retain that cheaper route when it suffices; reconstruct dependencies when adaptation or explanation requires them. Sections :4.1–:4.4 make that choice explicit. The source provides a discrete active-inference account and a minimal implementation, not evidence that one learning sequence suits every practice or agent.

How can notation reveal what an application still lacks? Adapt SICM’s Preface: recover function arguments, substitutions and composition where implicit conventions obscure the operation. A familiar conventional expression is cheaper when the reader can already use it reliably. Sections :4.2–:4.4 therefore call for operative reconstruction at the unresolved step; :5.2 supplies a case. The selected contribution is explicit operation recovery. Its source’s strong requirement of automatic interpretability is appropriate for computational exposition; ordinary use may be settled by a clear hand-worked construction.

What does a theory’s formal structure contribute to an application? Adopt Baez and Stay, §2.3’s constructive comparison of cartesian and monoidal operations. A familiar “parallel composition” description can leave input duplication unresolved. Section :5.1 obtains the required operation or the obstruction under stated rules, then asks whether the receiving practice can supply it. This pays the cost of recovering only the consequential structure. The broader cross-disciplinary correspondences in that paper do not establish the physical copying operation in our case.

The common method is a synthesis of these contributions. Reopen it when a better approach obtains a usable application with less reconstruction, when a recurring failure requires another connecting operation, or when changed tools alter the contribution the receiver must understand.

B.5.TU:12 - Relations

• B.5 coordinates the inquiry and a changed question. B.5.FM constructs a first model when consequential participants or relations remain unclear.
• B.5.RC, B.5.RA and B.5.RR supply construction recovery, argument recovery and reasoning revision.
• B.5.4 constructs a situational interpretation of an available concept. C.29 and C.29.1 supply mathematical correspondence and result transfer.
• C.29.2, C.29.3 and B.5.MPC connect computation, realization and the physical question.
• A.6.3.RT prepares an operative expression; C.2.8 helps characterize what a recipient can extract from an explanation under stated preparation and access.
• A.15.9 supplies bounded professional help; C.11.DUA selects useful further effort; C.39 develops a missing way.
• C.40 develops questions and ways together when the application opens that further work.

B.5.TU:End