# Connecting design theory and sociology: where does novelty get lost?

- Author: Shuichiro Ogawa (Design Researcher / Design Consultant), https://jamjet-iqbal.com/
- Published: 2026-07-27; updated: 2026-07-27
- Canonical (HTML): https://jamjet-iqbal.com/en/notes/design-sociology-research-questions/
- Japanese version: https://jamjet-iqbal.com/notes/design-sociology-research-questions/ (Markdown: https://jamjet-iqbal.com/notes/design-sociology-research-questions.md)
- Language: en
- Keywords: design-theory, sociology, research-questions, novelty, problem-definition, abduction, symmetry-inversion, evaluation, generative-AI

> To derive research questions from the areas where design theory and sociology have not been connected, we ran three generators with different methods in parallel and used their degree of convergence as an indirect indicator of novelty. Checking the primary literature and a second review by an academic critic rejected this indicator in two steps. First, the three generators shared the input corpus, so convergence measured salience rather than the independence of methods. Second, the candidates that failed all failed for the same reason, regardless of their degree of convergence: the general theories of sociology already included design practice within their scope. We made the demotion norm explicit in three stages and rescored the 15 candidates. The result was 10 of the covering type, 1 pending, 3 that stand after rewriting, and 2 of the symmetry-completion type (1 of which carries the reactivity of self-application). The note records, as a ledger, 15 integrated RQ candidates, 4 reverse contributions, 3 critiques of the author's own theory, and 52 unverified items.

## We assumed that convergence could stand in for novelty

To derive research questions from the areas where design theory and sociology have not been connected, we set up three generators and ran them in parallel.
The generators did not see each other's outputs, and each was given a different set of norms to refer to.
The three generators were as follows.

- A generator that invents whole frames through second-order abduction (see [Where Are the Gaps in AI Research? Seven Voids Found by Reframing](https://jamjet-iqbal.com/en/notes/ai-research-gaps-abduction/) and [Thinking in Problem Definition: Rereading the Game Board of Existing Research](https://jamjet-iqbal.com/en/notes/problem-definition-metagame/))
- A generator allowed only to reverse asymmetries with the A–G sheet ([AI Research Gaps, Revisited: Sorting Symmetry-Completion from Coverage, and Digging Out an Endogenous Scaling Law](https://jamjet-iqbal.com/en/notes/ai-research-gaps-revisited/), symmetry-inversion-method-assessment (private))
- A generator allowed only to build structural correspondences from unconnected sociological theories to design theory ([The Nexus of Design and Social Science: A Genealogy of Methods, Theories, and Institutions](https://jamjet-iqbal.com/en/notes/design-social-science-nexus/), litmap-meritocratic-evaluation (private))

If three generators with different methods fall into the same gap, that gap does not depend on method.
If they do not, the gap is only a habit of one method.
On this reasoning, we used the degree of convergence as an indirect indicator of novelty.

The generators produced 31 candidates, which converged into 8 clusters.
Four clusters had a convergence of 3, one had a convergence of 2, and three had a convergence of 1.
When we looked at this distribution, it seemed rational to start writing papers from the clusters with a convergence of 3.

When we had an academic critic (canon: design-theory) examine the result, this rationality was the first thing to break.

## The three generators were not independent

The critic's point was at the level of methodology.
The norm notes that the three generators referred to were different, but they shared the input corpus.
Moreover, the corpus contained only 14 unconnected sociological theories for the first generator, and these 14 had already been selected by the criterion of "not yet connected to design research."
If three generators choose from the same 14 theories, convergence measures what the inputs have in common rather than the independence of the methods.

The critic also pointed to an effect of salience.
The first cluster with a convergence of 3 (C1) took Raj, Berg, and Seamans's study of the AI disclosure penalty as the starting point for all three generators.
With 16 preregistered experiments and N=27,491, it is the most conspicuous empirical study in the corpus.
That all three started from the same place can be explained as a function of salience rather than as agreement among methods.

These two points removed the grounds for reading the degree of convergence as an order of priority.
What remained was to rescore the candidates one by one against the conditions of novelty.

## Which conditions did we use to score novelty?

We used the A–G sheet (symmetry-completion type) established in earlier work without change.

- **A**: State the common view to be reversed in a form that can be identified as a doctrine
- **B**: State the operation that reverses an asymmetry into a symmetry
- **C**: State a structural correspondence from another field
- **D**: State a one-to-one correspondence across categories
- **E**: State a prediction about empirical conditions
- **F**: State a test that can refute that prediction
- **G**: State the minimal observable, down to its unit

If D or G cannot be written, the candidate is not of the symmetry-completion type but of the **covering type**.
Work of the covering type extends a known frame to a new object; it has value, but it cannot claim a theoretical contribution.

We fixed the problem statement to the form "Existing research has explained Y on the assumption X. However, that assumption cannot identify Z."
A candidate that cannot name the assumption X has not yet written what it reverses.

## The demotion norm had only one sentence: "D cannot be written"

After the first judgment, the critic examined the consistency of the judgment itself.
What this exposed was that the demotion norm lacked a substantive criterion.
The norm said that a candidate is of the covering type if D cannot be written, but it did not say how to judge whether D had been written.

What had actually happened?
The candidate that moved Hirschman's exit, voice, and loyalty to the design profession was demoted on the grounds that "D cannot be written," whereas RQ-9, which moved Tilly's four mechanisms to design organizations, was kept.
The only substantive reason that separated the two was that Tilly's D column had five rows and looked concrete.
That is not the application of a rule; it is a proxy based on appearance.

The demotion norm that we made explicit has three stages.

1. Is the target outside the scope that the base theory explicitly claims? If it is inside, the candidate is of the covering type (D does not hold). **The number of rows and the concreteness of the D column are not considered.**
2. If it is outside, count how many terms of the base relations map. If not all terms map, the one-to-one correspondence fails and the candidate is of the covering type.
3. If some terms do not map, check whether that failure to map can be turned into an empirical claim. If it can, place the failure to map, rather than the mapping, at the core of the question.

The difference from before is that stage 1 is an independent gate.
If the base theory already includes the target in its scope, then however tightly the correspondence table is written, it is not an analogy but merely the application of an abstraction.

## The candidates failed on scope, not on convergence

Applying this norm uniformly to the 15 candidates substantially rewrote the first judgment.

Stage 1 alone removed nine candidates.
Fourcade and Healy's theory of classification includes the rating of labor markets within its scope, so RQ-1 is inside.
Zelizer's relational work includes the reward practices of creative labor.
Luhmann's second-order observation is a theory of evaluation in general, and design evaluation is inside it.
The same holds for Boltanski and Thévenot's model of orders of worth, Kornberger and colleagues' theory of evaluative infrastructures, and Becker's theory of conventions; each already has design practice within its range.
Tilly's *Durable Inequality*, according to its abstract, treats the organizations that produce categorical inequality as ranging from something as small as a household to something as large as a government.

The first judgment read the result "not one candidate survived from the four clusters with a convergence of 3" as a failure of the convergence indicator.
That reading was wrong.
**All the failed candidates failed for the same reason: they applied a general theory of sociology to an object already inside its scope.**
That the failed candidates came from the high-convergence clusters was a coincidence, and candidates from the low-convergence clusters failed at the same gate.

The assumption that collapsed lies in the A–G sheet itself.
The sheet implicitly assumed that "sociological theory and design theory are separate fields, and a mapping between them is an analogy."
The general theories, however, were written from the outset with a scope that includes design.

## The two survivors are not of equal standing

The two candidates that remained of the symmetry-completion type are RQ-11 and RQ-12, and both belong to the cluster with a convergence of 2 (C5).

**RQ-11**: In the co-evolution of design, which comes first in time when the problem space is redefined, the designer's cognitive updating or the updating of external evaluative infrastructure?

Existing research has assumed that the mutual updating of problem and solution spaces is a process within the designer's cognition (Crilly's two-part study, Gero and colleagues' formalization, Cash and colleagues' cognitive co-evolution model).
That assumption cannot identify cases in which approval gates, the constraints of a design system, or metric dashboards initiate the redefinition of the problem space first.
The minimal observable is the difference between the time at which references to the problem space changed and the times of approval records, design system commits, and changes to metric definitions (in time, with sign).
Cash and colleagues (2023) state in their abstract that there is a "critical disconnect" between the understanding of how designers think and the description of how they act.
This question fits directly into that disconnect.

**RQ-12**: In design organizations after the introduction of generative AI, which quantities have newly become measurable, and how have those quantities reshaped designers' allocation of work?

Existing research has assumed that generative AI is a tool fed into the process and has explained changes in productivity and creativity.
That assumption cannot identify the reactivity by which the introduction of AI converts "design outcomes" into newly commensurable quantities (the number of proposals, the time to a first draft, the approval rate) and practice itself is reshaped to fit those quantities.
Espeland and Sauder identified two mechanisms that produce reactivity: self-fulfilling prophecy and commensuration.
Of these, the path from commensuration to reactivity maps in a directly testable form.
The minimal observable is the year-on-year rate of change in the semantic dispersion of submitted proposals per review (the mean pairwise distance between the embedding vectors of the proposals).

The second review withdrew the scoring that placed these two candidates side by side.
**RQ-12 takes the reactivity of measurement as its object, yet it is itself a measurement of the fourth order and is subject to the same reactivity.**
If an apparatus that measures the semantic dispersion of submitted proposals is introduced into reviews, the apparatus becomes a new commensurable quantity and reshapes practice.
RQ-12 becomes an instance of the mechanism that it describes.

This is not fatal, but the scoring conditions differ from those of RQ-11.
The time differences of RQ-11 can be taken from quantities on the side of the first and second orders (the times of utterances and of commits), so the problem of performativity is small.
The first-round conclusion that "only RQ-11 and RQ-12 are theoretical contributions" was the product of scoring that ignored this difference.

## Some candidates come back when rewritten

For the three candidates that passed stage 3, leaving them demoted was a mistake.

**RQ-4 (orders of justification)** is confirmed as being of the covering type.
However, one point remains.
In the critic's reading, worth in Boltanski and Thévenot's model is a state that is re-established at each test.
If worth in design evaluation is attributed permanently to persons, that is not a failure of correspondence; it is a finding that design evaluation does not satisfy the axioms of the model.
The question can be rebuilt as "Do justifying utterances in design reviews attribute worth to the artifact now under test, or to the proposer's past record?"
The proportion (%) of proposers who received a negative judgment in the previous test and were able to submit a proposal to the next test can be obtained from submission records.



**RQ-7 (records of rejected proposals)** is the candidate for which the first round said, "replace the receiving theory, Frickel's undone science, with Schön's back-talk."
That instruction was itself incomplete.
Schön lies inside design theory, so if the receiving theory is Schön, column C (a structural correspondence from another field) cannot be filled by definition.
If the base is Schön's virtual world and the target is a generative AI session, the mapping stays within the field but crosses media.
For A, we place Simon's cycle of generation and testing and his satisficing.
The doctrine that pruned branches of search are a residue that the evaluation function correctly discards is reversed from the side that holds that losing the series of co-evolution means losing the process.

The core of the mapping is the switch of the acting agent ("the designer made a move" ↔ "the model made a move"), and only here does an antisymmetry arise that has no counterpart in the base.
The observables are the proportion of proposals for which the reasons for rejection can be stated afterward (%, stratified into self-made and generated proposals) and the median dwell time from the presentation of a proposal to its rejection (seconds per proposal).

Under a strict application of the ledger's scoring rule, RQ-7 is of the covering type, because the mapping stays within the field.
This result does not come from a weakness of the candidate. **It comes from a scoring rule that tries to judge two kinds of novelty (a symmetry reversal within a field and a structural mapping between fields) with the single column D.**
The reversal of Simon against Schön is a legitimate reversal that occurs inside design theory, and a rule that scores it as "covering type because it does not cross fields" mechanically closes the route of building questions from debates within the canon.
What needs to be fixed is the scoring rule.

The critic also named the places that could not be closed.
"Not being recorded means that the series of responses from the situation is cut off" cannot be derived from Schön alone.
Back-talk is received immediately within action, so even if a proposal is deleted, the response before deletion has already been received.
To make the proposition hold, an additional premise is needed: "under generative AI, proposals are produced faster than back-talk can be received."
Moreover, a repertoire is the tacit store of an individual practitioner, and identifying it with the set of recorded rejected proposals requires the very operation that Schön's own critique of technical rationality rejected.

A normative reservation was also attached.
RQ-7 treats "not being recorded" uniformly as a loss, but a history of who rejected what can be used directly as evidence of capability.
The question is not "recording is good" but "who can access the records."

**RQ-14 (is design a function system with its own code?)** stands if the question is replaced.
The first judgment (that the rate of reversed judgments when economic consequences are made explicit only measures economic subordination) was right in its conclusion but shallow in its reason.
The real reason that the critic gave is that, in Luhmann's system, structural coupling presupposes operational closure.
Strong coupling is not evidence against having a proper code; on the contrary, being able to speak of coupling at all presupposes closure.

As a check of the primary source, the critic also pointed out a more basic error.
According to the critic, the primary code of Luhmann's art system is **art / non-art**, and beautiful/ugly and original/imitative belong to **programs** (rules that apply the code to concrete contexts).
Programs are variable and may import criteria from other systems, so "judging on the basis of ROI" does not mean that the code belongs to the economy.

Three observables can be written for the discrimination.

- **Selectivity**: the proportion of variance explained by nonlinear terms in the rate of change of judgments when the size of economic consequences is manipulated at five levels
- **Recursive connection**: the explained proportion of judgment variance when only the series of preceding design judgments is changed, with the same artifact and the same economic information
- **Availability of the negative value**: the proportion of negative judgments maintained only by justifications that contain no economic vocabulary

Of these, recursive connection carries the most information.

However, what could be written is a different quantity.
All three measure the degree of operational closure, not whether a proper code exists.
Luhmann's codes are binary and categorical, so however precisely a continuous quantity is measured, it does not answer a categorical question.
This is not a failure in designing G but a mismatch between how the question is posed and the syntax of the theory.
If the question is rebuilt as "To what degree does the communication of design connect recursively to its own preceding operations?", observables can be written. However, that is the operationalization of a peripheral concept rather than a test of the theory of differentiation, and the theoretical contribution becomes correspondingly smaller.

## Only one judgment was left pending

RQ-10 (are design systems devices that replicate categorical pairs?) has the same structure as RQ-9 in its use of Tilly, but it was not demoted.

In the critic's reading, Tilly's emulation and adaptation are formulated as mechanisms carried by people and by organizational models.
What RQ-10 claims is that the carrier is an **artifact**, a published document of conventions, and that this artifact is at the same time a **tool of production**.
We have not confirmed whether cases in which the mechanism that reproduces inequality and the tool of production share the same substrate fall within Tilly's descriptions.

There is one thing that can be added from the side of design theory.
This is a position where Buchanan's second order (things) and fourth order (environments) overlap on the same object.
There, the character of Schön's virtual world as "a place where moves are made" and the character of a governance document as "deciding who can make moves" coexist in the same thing.
This coexistence may have no counterpart on Tilly's side.

There is only one condition for lifting the pending judgment.
It is to check whether *Durable Inequality* contains descriptions that treat technical artifacts as carriers of the mechanisms.
If it does, RQ-10 is demoted like RQ-9; if it does not, RQ-10 is re-examined as being of the symmetry-completion type.

RQ-9, on the other hand, was demoted.
The ledger has two possible rebuttals, and neither holds.
"The pair of having and not having AI authority is a new pair that is not among Tilly's cases" does not hold.
Emulation is precisely the mechanism by which existing categorical distinctions are transplanted between organizations, and the generation of new pairs is inside the theory.
"Opportunities to explore are non-zero-sum goods, so the correspondence differs from opportunity hoarding, which assumes finite resources" does not hold either.
This is the ledger's own claim in reverse contribution 1, but if it holds, it breaks D for RQ-9.
The way to reconcile them is to restate what is hoarded as the allocation of time, place, attention, and opportunities for repetition, but these are finite resources, so the case returns directly to Tilly's standard cases.

The reason for demoting RQ-8 was also corrected.
"A proposal has no place to exit to, so the asymmetry between leaving and re-entering does not map" is not a problem of scope.
Moving from the relation between an actor and an organization to the relation between an actor and their own work is a legitimate crossing of categories, from a relation between actors to a relation within an actor.
What fails is not the crossing but the one-to-one correspondence: of the three terms, exit lacks a destination and loyalty has no counterpart.
If the counterpart of loyalty is rebuilt as "the time spent continuing to watch the consequences of a move," RQ-8 could return by sharing the same observable (dwell time) as the rewritten RQ-7.

## The scoring sheet itself had four defects

The critique was directed at the sheet as well as at the candidates.

First, **Buchanan's four orders do not partition the 15 candidates**.
The first judgment carried the reservation that "there is no guarantee that the 15 candidates can be scored side by side on the same sheet, because they belong to different orders."
When we actually assigned them, two could not be assigned (RQ-1, whose object is hiring selection rather than design activity, and RQ-15, a normative meta-problem), and four were mixed (RQ-7, RQ-10, RQ-11, RQ-13).
"A separate sheet for each order" cannot be operated.

Second, **G divides into three types**.
The basis of the division is not the order but where the unit of observation comes from.
The type in which the unit is given by the boundary of the artifact (G type I) can stay as it is.
The type in which the unit is created by the analyst's coding scheme (G type II) needs a column that shows that the scheme can be defined independently of the hypothesis being tested.
The type in which the boundary of the object itself is not given (G type III) needs a column for the rule of boundary setting and for an upper bound on the influence that introducing the measuring apparatus has on the object.
Among the current 15, five are of G type III (RQ-5, RQ-6, RQ-9, RQ-12, RQ-14), and in all of them the column on the self-application of reactivity is empty.
Four are of G type II (RQ-3, RQ-4, RQ-7, RQ-13), and in all of them the theory-independence of the unit is not filled in.

Third, **the doctrine reversed in A must be a doctrine about the order that the RQ targets**.
Simon's theory of search is a doctrine about problem solving in the first and second orders, so using it for A in an RQ of the fourth order is an error of scope.
Under this rule, the A columns of RQ-9 and RQ-12 need to be rewritten.

Fourth, **in 11 of the 15 candidates, the F column is merely the mechanical negation of the E column**.
A refutation condition must state what would be observed if the rival hypothesis were true.
"If the prediction of E is not observed, the reversal is rejected" does not name a rival hypothesis.
The only exception is RQ-12, whose F names a rival hypothesis (a mechanism on the generation side).

## The list of integrated RQ candidates

The table records all 15 candidates in the ledger, with the second-round rejudgment.
The detailed A–G sheets are in `source/review/design-sociology-blindspots/rq-candidates.md`.

| # | Question | Sociology used | Minimal observable | Second-round rejudgment |
|---|---|---|---|---|
| RQ-1 | Does the AI disclosure penalty transfer from the work to the author? | Fourcade and Healy | Difference in pass rates of document screening between groups with and without a disclosure history (%pt) | Covering type (the theory of classification includes the rating of labor markets) |
| RQ-2 | Is the penalty determined by the type of relationship rather than by authenticity? | Zelizer | Acceptance rate for each cell of relationship type × process (%) | Covering type (Alacovska and colleagues have already done this). Highest feasibility |
| RQ-3 | Do evaluators look at the work, or at the eyes of other evaluators? | Luhmann | Gap between first-order and second-order evaluations | Covering type (second-order observation is a theory of evaluation in general) |
| RQ-4 | Who controls the switching of orders of justification? | Boltanski and Thévenot | Distribution of order labels per utterance and number of switches | Covering type. Rebuild only the failure to map R4 (the reversibility of worth) |
| RQ-5 | Who holds the authority to revise the evaluation axes? | Kornberger; Justesen and Plesner | Number of criterion revisions and the share originating from those evaluated (%) | Covering type (the theory of evaluative infrastructures includes the institutions of criterion revision) |
| RQ-6 | Which order's tests does generative AI automate? | Boltanski and Thévenot | Number of AI applications by order (count per order) | Covering type (the formalizability of tests is inside Boltanski and Thévenot) |
| RQ-7 | How does generative AI change the retention rate of rejected proposals? | Schön's virtual world (receiving theory replaced) | Proportion of proposals whose reasons for rejection can be stated afterward (%), dwell time (seconds per proposal) | Point of dispute. The mapping is within the field, but the scoring rule should be fixed |
| RQ-8 | When switching between proposals becomes cheap, does staying die? | Hirschman | Number of edits per proposal, dwell time, revisit rate | Covering type (absorbed into RQ-7). Reason for demotion corrected to a failure of one-to-one correspondence |
| RQ-9 | Along which categorical pairs are opportunities to explore unevenly distributed? | Tilly | Gini coefficient across job levels of the allocation of time to undetermined tasks | **Demoted to covering type** (first-round judgment withdrawn) |
| RQ-10 | Are design systems devices that replicate categorical pairs? | Tilly | Similarity of wording of governance documents across firms | **Pending** (whether cases with an artifact as carrier exist is unconfirmed) |
| RQ-11 | Is co-evolution driven by cognition or by evaluative infrastructure? | Kornberger; Espeland and Sauder | Time difference between cognitive updating and infrastructure updating (signed) | **Symmetry-completion type** (maintained) |
| RQ-12 | Newly measurable quantities reshape practice | Espeland and Sauder | Year-on-year rate of change in the semantic dispersion of submitted proposals (% per year) | **Symmetry-completion type**. However, G is performative and the scoring conditions differ |
| RQ-13 | Are the agents of homogenization neither individuals nor groups but shared conventions? | Becker | Elasticity of proposal dispersion with respect to the number of shared conventions (% per convention) | Covering type (the theory of conventions is a theory of collective production in general) |
| RQ-14 | Is design a function system with its own code? | Luhmann | Proportion of judgment variance explained by the series of preceding judgments (%) | Stands if the question is rebuilt (toward the degree of recursive connection) |
| RQ-15 | How should inequality in non-zero-sum goods be measured? | Tilly; the limits of the scope of Fourcade and Healy | Comparison of the rate at which exploration succeeds across the conditions that make it possible | Covering type (the doctrine reversed in A cannot be identified) |

## Reverse contributions from design to sociology

Applying structural correspondences left four residues on the sociological side.
They are positions that could become the core of a paper's originality, but for none of them has it been confirmed that the structure is "absent on the sociological side."

1. **Inequality in non-zero-sum goods**: Tilly, and Fourcade and Healy, assume positional goods or finite resources. Rankings are zero-sum by definition. A proposition can be stated that inequality in goods that do not decrease for others when they increase, such as the possibility of exploration, arises not in the goods themselves but in the conditions that make exploration possible.
2. **Re-entry after exit**: Hirschman's three types close with exit and do not deal with the conditions under which those who have left return. The earlier review judged that the 1970 edition contains no discussion of re-entry (a full-text check against the source has not been done). A design for re-entry could add a fourth term.
3. **Ranking when the objects of classification increase endogenously**: Fourcade and Healy describe the falling cost of classification as an expansion of ranking, but they treat the objects of classification as exogenous. In design, the cost of generation falls at the same time, so the number of objects to be ranked increases endogenously.
4. **Externalizing blind spots through making**: Luhmann's second-order observation is an operation of observation and does not include making. Design externalizes the blind spots of observation not through description but through the making of artifacts.

The critic pointed out a problem of attribution in these four.
Items 1 and 2 are claims of single-author theories, not shared assets of the field of design research.
To claim the direction "from design to sociology," it must be shown that the structure is established on the side of the field.
Items 3 and 4 pass as assets of the field, but their absence on the sociological side remains to be checked.



## The order of writing papers

In the lane of theoretical contribution, RQ-11 stands alone at the front.
RQ-12 is in the same cluster, but because it is subject to the reactivity that it describes, the treatment of performativity must be handled in the text.

As candidates that return after rewriting, RQ-7 and RQ-14 follow.
For both, the observables have been written, and what remains is a gap in the argument (leap 2 of RQ-7) or the rebuilding of the question (RQ-14).

For RQ-10, the judgment is settled by checking one primary source.
If *Durable Inequality* does not treat technical artifacts as carriers of the mechanisms, RQ-10 goes up for re-examination as being of the symmetry-completion type.

In the lane of feasibility, the order changes.
RQ-2 is complete if the type of relationship is crossed orthogonally with the experimental design of Raj and colleagues as a second factor.
It fell into the covering type, but a single preregistered experiment decides whether the reduction to a single mechanism is correct.

We do not perform the operation of mixing the two lanes and choosing a single "best candidate."
Lining up candidates judged by different criteria in a single row was the cause of the failure of convergence.

## Conclusions not to adopt

- "Because Boltanski and Thévenot's seven orders map onto the vocabulary of design evaluation, design evaluation can be explained as a clash of multiple orders of justification." A match in vocabulary is a redescription, not an explanation. The more smoothly the seven terms are filled, the stronger the mapping looks, so the correspondence table that most deserves suspicion is the one most likely to be adopted.
- The first-round conclusion that "only RQ-11 and RQ-12 are theoretical contributions." The two have different scoring conditions.

## Unresolved conflicts

| Axis of conflict | Position A | Position B | Evidence needed to decide |
|---|---|---|---|
| Scoring criterion for the Boltanski and Thévenot correspondence table | Score it as a mapping of relations (Gentner 1983) | Score it as the creation of a frame (Dorst 2011) | Whether Gentner 1983 states the constraints of one-to-one correspondence and parallel connectivity explicitly |
| Conditions for receiving back-talk | Received immediately within action (Schön 1983) | Constructing a repertoire requires records (Argyris and Schön) | Whether the section on repertoire in *The Reflective Practitioner* requires records in the formation of the store |
| Is not recording rejected proposals a loss? | It is the normal operation of satisficing and not a loss (Simon 1969) | Losing the series of co-evolution means losing the process (Dorst and Cross 2001) | Whether Simon's pruning of the search tree and Dorst and Cross's series of problem-solution pairs refer to the same object |
| Measurability of a proper code | Binary and categorical, so it cannot be measured with continuous quantities | It can be measured continuously as a rate of recursive connection | Whether the theory of code in *Art as a Social System* permits a reading by degree |
| Partitioning the scoring by the four orders | They line up by the indeterminacy of subject matter, so they can be used to partition (the first-round judgment) | They are placements, not separate compartments (Buchanan 1992) | In the text, Buchanan 1992 states that the four areas are "also interconnected, with no priority given to any single one," which supports position B |
| Is the application of Tilly an extension or an instantiation? | Design organizations are inside the scope (instantiation) | Outside the scope, because an artifact becomes the carrier (extension) | Whether *Durable Inequality* contains descriptions that treat technical artifacts as carriers of the mechanisms |
| Norms for record-keeping institutions | Records of rejected proposals should be increased as a learning resource | Records put proposers at risk; the design of access rights comes first (Costanza-Chock 2020) | How the relevant chapter of *Design Justice* treats archives and access rights |

## Methodological limitations

In this study, we were not able to carry out the step of checking the comparison of options against an external second opinion.
Codex had reached its credit limit, and the next date on which it would be available was 2 August 2026.
As a substitute, the academic critic served as the second opinion and in fact rejected major design decisions twice (the ordering by convergence and the asymmetric application of the demotion norm).
However, the critic read the same corpus, so it did not provide an independent check of the selection of the corpus itself.

The other limitation is the proportion of the primary literature relied on that has not been checked.
Many of the second-round judgments were made without checking the relevant passages of the original sources directly.
Ten primary sources have been designated as the next to read, and one of them, *Durable Inequality*, alone decides the judgment on RQ-10.

## Unverified items

Combining the 28 items in the ledger and the 25 items from the critic and removing duplicates gives 52 items.
Only the main points are listed below; all items are in the section "Unverified markers (aggregated)" of `source/review/design-sociology-blindspots/rq-candidates.md`.

### Resolved in the second-round rejudgment (9 items)

- The complete bibliographic record of Rughinis, Vulpe, and Turcanu 2025 (CSCS 2025, pp.572–579, DOI 10.1109/CSCS66924.2025.00091)
- Peer-reviewed publication of van der Maden and colleagues 2025 (DIS '25 Companion, DOI 10.1145/3715668.3734170)
- Tilly's mechanisms confirmed as four (exploitation, opportunity hoarding, emulation, adaptation)
- Zelizer's definition of earmarking (in addition to the differentiation of meanings, it includes normative distinctions and the adoption of specific media of exchange)
- That Hirschman's three types do not include re-entry
- The bibliographic record of Foucault 1977 (Pantheon Books, ISBN 978-0394499420; for books, the ISBN and WorldCat serve as the stable record)
- The bibliographic record of Granovetter 1985 (*AJS* 91(3), 481–510, DOI 10.1086/228311)
- The size of the change in Doshi and Hauser 2024 (the increase in similarity between stories as a share of the range of similarity in the human-only condition: 10.7% for the group with one AI idea and 8.9% for the group with five)
- The sample of Moon and colleagues 2025 (2,200 college admissions essays) and the condition in Sun and colleagues 2025 (the models that reached the 52nd percentile were Claude and GPT-4)

### Corrected (3 items)

- According to the critic's check of the primary source, **Luhmann's primary code is art / non-art**, and beautiful/ugly belongs to programs. The passages that wrote "beautiful/ugly = proper code" were wrong
- **The term "stratified adoption" could not be found in Woronkowicz and colleagues 2026**. The paper is a special-issue editorial, and what may be reported is limited to the 2025 SNAAP Pulse survey (more than 2,000 arts and design alumni)
- **The claim that Zelizer is absent collapses**. Alacovska 2018 applies the relational work perspective to the informal labor practices of creative workers, and Alacovska and colleagues 2024 apply it to interviews with 49 visual artists, graphic designers, and illustrators. A safe formulation is "relational work has been adopted in research on creative labor on platforms, but how the practice of AI disclosure intersects with types of commissioning relationship has not been examined"

### Unresolved items that affect the judgments (excerpt)

- `[unverified: full text not reached, abstract only]` Whether Tilly 1998 contains descriptions that treat technical artifacts as carriers of the mechanisms (the only condition for lifting the pending judgment on RQ-10)
- `[unverified: full text not reached, abstract only]` The formulation that Luhmann's structural coupling presupposes operational closure (the basis of the judgment on RQ-14)
- `[unverified: not attempted, the search of secondary literature is not finished]` Whether there is an interpretation that reads Luhmann's system differentiation by degree (if there is, the judgment on RQ-14 collapses)
- `[unverified: full text not reached, the text of On Justification not obtained]` The axioms of Boltanski and Thévenot's model and the formulation that worth is re-established at each test (the core of the rewritten RQ-4)
- `[unverified: full text not reached, abstract only]` Whether Gentner 1983 includes one-to-one mapping and parallel connectivity, or whether these are formulations from SME 1989 onward (the attribution of the scoring criterion changes)
- `[unverified: full text not reached, the text of the book not obtained]` Whether Schön 1983's repertoire is limited to the tacit knowledge of individuals or includes collective records (leap 2 of RQ-7)
- `[unverified: not obtainable, three routes (DOI, publisher, Wayback)]` Whether Schön and Wiggins 1992 place a lower time limit on see-move-see (G2 of RQ-7)
- `[unverified: bibliographic details not settled, chapter pages and edition]` The chapters, pages, and edition of Simon 1969 on generator-test and satisficing (A of the rewritten RQ-7)
- The reading that Buchanan 1992's four orders line up by the indeterminacy of subject matter: in a check against the text, Buchanan states that the four areas are "also interconnected, with no priority given to any single one," so the original does not support this reading



- `[unverified: full text not reached, publisher's bibliographic page only]` Costanza-Chock 2020's position on a community's veto (the normative reservation on RQ-7)
- `[unverified: full text not reached, NordiCHI '22]` Whether NordiCHI '22 "The Troubling Cups" uses Tilly (a check of the text of Sloane 2019 found no citation of Tilly)
- `[unverified: full text not reached, summary page only]` The identification strategy of Brynjolfsson and colleagues 2025



## Related notes

The material and the judgments on the state of application are in design-sociology-blindspots-literature (private).
The conditions of novelty and the method of Problem Definition are defined in symmetry-inversion-method-assessment (private), [Thinking in Problem Definition: Rereading the Game Board of Existing Research](https://jamjet-iqbal.com/en/notes/problem-definition-metagame/), and [AI Research Gaps, Revisited: Sorting Symmetry-Completion from Coverage, and Digging Out an Endogenous Scaling Law](https://jamjet-iqbal.com/en/notes/ai-research-gaps-revisited/).



## References

### Design theory canon used in the judgments

- Buchanan, R. (1992). Wicked Problems in Design Thinking. *Design Issues* 8(2), 5–21. https://doi.org/10.2307/1511637
- Buchanan, R. (2001). Design Research and the New Learning. *Design Issues* 17(4), 3–23. https://doi.org/10.1162/07479360152681056
- Simon, H. A. (1969). *The Sciences of the Artificial*. MIT Press. `[unverified: bibliographic details not settled, chapter pages and edition]`
- Schön, D. A. (1983). *The Reflective Practitioner: How Professionals Think in Action*. Basic Books.
- Schön, D. A. & Wiggins, G. (1992). Kinds of seeing and their functions in designing. *Design Studies* 13(2), 135–156. https://doi.org/10.1016/0142-694X(92)90268-F
- Rittel, H. W. J. & Webber, M. M. (1973). Dilemmas in a General Theory of Planning. *Policy Sciences* 4(2), 155–169. https://doi.org/10.1007/BF01405730
- Cross, N. (1982). Designerly ways of knowing. *Design Studies* 3(4), 221–227. https://doi.org/10.1016/0142-694X(82)90040-0
- Dorst, K. & Cross, N. (2001). Creativity in the design process: co-evolution of problem–solution. *Design Studies* 22(5), 425–437. https://doi.org/10.1016/S0142-694X(01)00009-6
- Dorst, K. (2011). The core of 'design thinking' and its application. *Design Studies* 32(6), 521–532. https://doi.org/10.1016/j.destud.2011.07.006
- Krippendorff, K. (1989). On the Essential Contexts of Artifacts. *Design Issues* 5(2), 9–39. https://doi.org/10.2307/1511512
- Krippendorff, K. (2006). *The Semantic Turn: A New Foundation for Design*. CRC Press.
- Kimbell, L. (2011). Rethinking Design Thinking: Part I. *Design and Culture* 3(3), 285–306. https://doi.org/10.2752/175470811X13071166525216
- Kimbell, L. (2012). Rethinking Design Thinking: Part II. *Design and Culture* 4(2), 129–148. https://doi.org/10.2752/175470812X13281948975413
- Costanza-Chock, S. (2020). *Design Justice: Community-Led Practices to Build the Worlds We Need*. MIT Press. https://mitpress.mit.edu/9780262043458/design-justice/
- Schumacher, P. (2011/2012). *The Autopoiesis of Architecture* I/II. Wiley. `[unverified: bibliographic details not settled, the formulation of a proper code]`

### Theories of analogy and structure mapping (basis of the scoring criteria)

- Gentner, D. (1983). Structure-Mapping: A Theoretical Framework for Analogy. *Cognitive Science* 7(2), 155–170. https://doi.org/10.1207/s15516709cog0702_3
- Falkenhainer, B., Forbus, K. D. & Gentner, D. (1989). The Structure-Mapping Engine. *Artificial Intelligence* 41(1), 1–63. https://doi.org/10.1016/0004-3702(89)90077-5
- Gentner, D. & Markman, A. B. (1997). Structure Mapping in Analogy and Similarity. *American Psychologist* 52(1), 45–56. https://doi.org/10.1037/0003-066X.52.1.45

### Sociological theories not yet connected to design

- Luhmann, N. (2000). *Art as a Social System*. Stanford University Press. https://doi.org/10.1515/9781503618763
- Becker, H. S. (1982). *Art Worlds*. University of California Press. https://academic.oup.com/sf/article-abstract/61/4/1268/1887386
- Boltanski, L. & Thévenot, L. (2006 [1991]). *On Justification: Economies of Worth*. Princeton University Press. ISBN 0-691-12516-3
- Boltanski, L. & Chiapello, È. (2005 [1999]). *The New Spirit of Capitalism*. Verso. https://www.versobooks.com/products/1980-the-new-spirit-of-capitalism
- Zelizer, V. A. (2011). *Economic Lives: How Culture Shapes the Economy*. Princeton University Press. https://press.princeton.edu/books/hardcover/9780691139364/economic-lives
- Tilly, C. (1998). *Durable Inequality*. University of California Press. https://doi.org/10.1525/9780520924222
- Hirschman, A. O. (1970). *Exit, Voice, and Loyalty*. Harvard University Press. https://www.hup.harvard.edu/books/9780674276604
- Fourcade, M. & Healy, K. (2013). Classification Situations. *Accounting, Organizations and Society* 38(8), 559–572. https://doi.org/10.1016/j.aos.2013.11.002
- Fourcade, M. & Healy, K. (2024). *The Ordinal Society*. Harvard University Press. https://doi.org/10.4159/9780674296688
- Foucault, M. (1977). *Discipline and Punish: The Birth of the Prison*. Pantheon Books. ISBN 978-0394499420
- Granovetter, M. (1985). Economic Action and Social Structure: The Problem of Embeddedness. *American Journal of Sociology* 91(3), 481–510. https://doi.org/10.1086/228311

### Prior applications of sociological theory to fields adjacent to design (counterexamples to claims of absence)

- Alacovska, A. (2018). Informal Creative Labour Practices: A Relational Work Perspective. *Human Relations* 71(12), 1563–1589. https://doi.org/10.1177/0018726718754991
- Alacovska, A., Bucher, E. & Fieseler, C. (2024). A Relational Work Perspective on the Gig Economy. *Work, Employment and Society* 38(1), 161–179. https://doi.org/10.1177/09500170221103146
- Alacovska, A., Bucher, E. & Fieseler, C. (2025). Algorithmic Paranoia. *New Technology, Work and Employment*. https://doi.org/10.1111/ntwe.12317
- MacDonald, T. W. L. (2026). Automating authenticity. *Economy and Society* 55(1). https://doi.org/10.1080/03085147.2025.2601464 `[unverified: full text not reached, citations of Zelizer and Tilly]`
- Jürgenmeyer, J. & Krenn, K. (2016). Classification Situations: A New Field of Research for Valuation Studies? *Valuation Studies* 4(2). https://doi.org/10.3384/VS.2001-5992.1642177
- Sloane, M. (2019). On the Need for Mapping Design Inequalities. *Design Issues* 35(4), 3–11. https://doi.org/10.1162/desi_a_00559

### Sociology of valuation, quantification, and work

- Espeland, W. N. & Sauder, M. (2007). Rankings and Reactivity. *American Journal of Sociology* 113(1), 1–40. https://doi.org/10.1086/517897
- Kornberger, M., Pflueger, D. & Mouritsen, J. (2017). Evaluative Infrastructures. *Accounting, Organizations and Society* 60, 79–95. https://doi.org/10.1016/j.aos.2017.05.002
- Mennicken, A. & Kornberger, M. (2021). Von Performativität zu Generativität. *KZfSS* 73(S1), 451–478. https://doi.org/10.1007/s11577-021-00755-x
- Kellogg, K. C., Valentine, M. A. & Christin, A. (2020). Algorithms at Work. *Academy of Management Annals* 14(1), 366–410. https://doi.org/10.5465/annals.2018.0174
- Justesen, L. & Plesner, U. (2024). Angry Citizens and Black Belt Employees. *Valuation Studies* 11(1), 9–37. https://doi.org/10.3384/VS.2001-5992.2024.11.1.9-37
- Frickel, S. et al. (2010). Undone Science. *Science, Technology, & Human Values* 35(4), 444–473. https://doi.org/10.1177/0162243909345836
- Woronkowicz, J., Noonan, D. S. & Skaggs, R. (2026). GenAI and Creative Labor: New Evidence on Valuation, Inequality, and Adoption. *Work and Occupations*. https://doi.org/10.1177/07308884261450882 `[unverified: full text not reached; a special-issue editorial, and "stratified adoption" was not found]`
- Rughinis, R., Vulpe, S. N. & Turcanu, D. (2025). Generative AI at the Crossroads of Values: Competing Orders of Worth in University and Publisher Regulations. *CSCS 2025*, 572–579. https://doi.org/10.1109/CSCS66924.2025.00091

### Recent developments and empirical studies in design theory

- Crilly, N. (2021a). The Evolution of "Co-evolution" (Part I). *She Ji* 7(3), 309–332. https://doi.org/10.1016/j.sheji.2021.07.003
- Crilly, N. (2021b). The Evolution of "Co-evolution" (Part II). *She Ji* 7(3), 333–355. https://doi.org/10.1016/j.sheji.2021.07.004
- Gero, J. S., Kannengiesser, U. & Crilly, N. (2022). Abstracting and formalising the design co-evolution model. *Design Science* 8, e14. https://doi.org/10.1017/dsj.2022.10
- Cash, P., Gonçalves, M. & Dorst, K. (2023). Method in their madness. *Design Studies* 88, 101219. https://doi.org/10.1016/j.destud.2023.101219
- Cash, P., Daalhuizen, J. & Cross, N. (2025). Designing: A new era for design research. *Designing* 1. https://journals.sagepub.com/doi/10.1177/30497671251392170
- Figoli, F. A., Bruggeman, R., Rampino, L. & Ciuccarelli, P. (2025). AI-against-design map. *Design Studies* 96, 101279. https://doi.org/10.1016/j.destud.2024.101279
- van der Maden, W. et al. (2025). When Discourse Stalls. *DIS '25 Companion*. https://doi.org/10.1145/3715668.3734170 (arXiv:2503.08565)
- Wadinambiarachchi, S., Kelly, R. M., Pareek, S., Zhou, Q. & Velloso, E. (2024). The Effects of Generative AI on Design Fixation and Divergent Thinking. *CHI '24*. https://doi.org/10.1145/3613904.3642919
- Doshi, A. R. & Hauser, O. P. (2024). Generative AI enhances individual creativity but reduces the collective diversity of novel content. *Science Advances* 10(28), eadn5290. https://doi.org/10.1126/sciadv.adn5290
- Moon, K., Green, A. E. & Kushlev, K. (2025). Homogenizing effect of large language models (LLMs) on creative diversity. *Computers in Human Behavior: Artificial Humans* 6, 100207. https://doi.org/10.1016/j.chbah.2025.100207
- Sun, L. et al. (2025). Large language models show both individual and collective creativity comparable to humans. *Thinking Skills and Creativity*. https://doi.org/10.1016/j.tsc.2025.101870
- Anderson, B. R., Shah, J. H. & Kreminski, M. (2024). Homogenization Effects of Large Language Models on Human Creative Ideation. *C&C '24*, 413–425. https://doi.org/10.1145/3635636.3656204
- Ashkinaze, J., Mendelsohn, J., Qiwei, L., Budak, C. & Gilbert, E. (2024/2025). How AI Ideas Affect the Creativity, Diversity, and Evolution of Human Ideas. *CI 2025*. https://doi.org/10.1145/3715928.3737481
- Raj, M., Berg, J. M. & Seamans, R. (2026). The Artificial Intelligence Disclosure Penalty. *Journal of Experimental Psychology: General* 155(4), 896–915. https://doi.org/10.1037/xge0001889
- "Better Ask for Forgiveness than Permission": AI Disclosure in Freelance Work. *CHI '26*. arXiv:2603.07459. https://arxiv.org/pdf/2603.07459
- Brynjolfsson, E., Chandar, B. & Chen, R. (2025). Canaries in the Coal Mine? Stanford Digital Economy Lab. https://digitaleconomy.stanford.edu/publication/canaries-in-the-coal-mine-six-facts-about-the-recent-employment-effects-of-artificial-intelligence/ `[unverified: full text not reached, identification strategy]`
- Shen, J. H. & Tamkin, A. (2026). How AI Impacts Skill Formation. arXiv:2601.20245. https://arxiv.org/abs/2601.20245
- Liu, G., Christian, B., Dumbalska, T., Bakker, M. A. & Dubey, R. (2026). AI Assistance Reduces Persistence and Hurts Independent Performance. arXiv:2604.04721. https://arxiv.org/abs/2604.04721