Systems, observers and levels
Second-order cybernetics, recursion, boundary critique: what already says state the level and no observer judges itself.
The seven lines of the Holagram way of thinking all have older statements, and the sources here show that what is new is the writing form, not the lines. 'No level sums up another' is argued by Cilliers (hierarchies are less nested than assumed) and by Subramanian and colleagues (the ecological and the individualistic fallacy); Morin's hologrammatic principle, which puts the whole inside the part, points the other way. 'No observer judges itself' is nearest to second-order cybernetics (von Foerster) and Luhmann's observation of observers, and in practice to Ulrich's boundary critique, which lets those affected but not involved contest the boundaries. The nearest method is Checkland's soft systems methodology: a model is a device for debate, tied to a declared worldview and set against the situation. What Holagram adds is narrower: a Trace anyone can open, a hands record per register, a labelled Fiction, a failure log and a rule against roll-up. The critics are Rittel and Webber (no stopping rule, no right to be wrong), Cilliers (a nested diagram is also a model), Kok and colleagues (complexity thinking lacks a theory of power) and the regress in second-order thinking, where the other observer sits in the same frame. For mandates, Luhmann's decision premises, Beer's rule that purpose is what a system does, and the Cybersyn case (Medina, Espejo) supply a reading frame.
Sources
- Gregory Bateson (foreword to the 2000 edition by Mary Catherine Bateson) (1972). Steps to an Ecology of Mind: Collected Essays in Anthropology, Psychiatry, Evolution, and Epistemology. Chandler (San Francisco), 1972; University of Chicago Press edition, 2000checkedmust read
Among these sources, the earliest statement of lines 1 and 5. The underlying point is that a message and a message about the message belong to different levels, and that confusing them produces pathology. Limit: logical types concern messages, so using 'level' for body, street and territory is an analogy. Italian route: Verso un'ecologia della mente (Adelphi, new expanded edition, 18th printing, 2001, 604 pages; Open Library record seen).
Read first: Part III: 'The Logical Categories of Learning and Communication' and 'Social Planning and the Concept of Deutero-Learning'. Part VI: 'Ecology and Flexibility in Urban Civilization'.
- Heinz von Foerster (2003). Understanding Understanding (essays including 'Cybernetics of Cybernetics' and 'Ethics and Second-Order Cybernetics'); earlier collection Observing Systems. Springer, New York, 2003 (Observing Systems: Intersystems Publications, 1981)checkedmust read
The nearest source for line 5. Von Foerster separated the cybernetics of observed systems from the cybernetics of observing systems, so that the observer belongs to what is described; Scott's abstract dates the distinction to 1974. For Holagram this means a Reading cannot be judged from inside its own frame, and the position of its author has to be written into the Entry. The same literature supplies the objection: the second observer is also an observer, so 'another level or another agent' does not leave the frame. Glanville, 'The purpose of second-order cybernetics' (Kybernetes 33(9/10): 1379-1386, 2004; Crossref record seen) is a shorter route in; its abstract says goal and purpose need rethinking within second-order systems, which bears on Beer's rule about purpose (thinking-04). Italian route: an Italian edition of Observing Systems, Sistemi che osservano (Astrolabio, 1987), is recalled and was not found in Open Library.
Read first: The chapters 'Ethics and Second-Order Cybernetics' and 'Cybernetics of Cybernetics' (titles only).
- Niklas Luhmann; edited by Dirk Baecker; translated by Rhodes Barrett (2018). Organization and Decision. Cambridge University Presschecked
For the mandates method, and for line 5. The blurb says Luhmann treats organisations as self-reproducing systems of decision communications, and the chapter titles name what a mandate document is made of: decision premises, decision programmes, personnel, and the organisation's own description of itself, with a chapter on the gap between reform and what the organisation does. Reading a statute or charter as a set of decision premises asks whether it fixes who may decide what, under which programme, which is close to 'can the unit take the next step'. Caution: Maturana, whose autopoiesis Luhmann extended to society, held that the concept applies only to the molecular domain (Wikipedia, fetched), so the borrowing is contested. Use the book for its questions about organisations, not as proof that a ministry is a self-producing system.
Read first: 'The Paradox of Decision-Making', 'Uncertainty Absorption', 'Decision Premises', 'Decision Programs', 'Self-Description' and 'Structural Change: The Poetry of Reform and the Reality of Evolution' (titles from the Cambridge contents list).
- Stafford Beer (2002). What is cybernetics? (an address delivered at the University of Valladolid, Spain). Kybernetes 31(2): 209-219checked
Gives the strategy and mandates methods two things. First, the heuristic that the purpose of a system is what it does: it turns 'strategy document against what the organisation does' into a rule, since a stated purpose is a Trace about the text and the purpose is read from outputs. Second, the viable system model's recursion: as the Wikipedia summary has it, any viable system contains and is contained in a viable system, with five functions at each level (operations, coordination, control, intelligence, policy). That is a working vocabulary for nested units such as federation, emirate, authority and free zone. Limit: the model is a design model with a control logic; applied to a mandate text it can read five boxes into any document.
Read first: The whole article (11 pages).
- Eden Medina (2011). Cybernetic Revolutionaries: Technology and Politics in Allende's Chile. MIT Presschecked
A history of Project Cybersyn, Chile's attempt to combine socialist reform under Allende with a computer system for managing the economy; the abstract says neither the reform nor the system was fully realised and that the book asks what the story shows about technology and politics. For Holagram it is the one government case in this thread in which a diagram of an institution was built, run and argued over, and its questions are those of the mandates method: who set the model's purpose, who could act on its signals, and what the reform text promised against what was done. Read it beside Espejo's first-person account (thinking-06).
Read first: 'Constructing the Liberty Machine' and 'Conclusion: Technology, Politics, History' (titles only).
- Raul Espejo (2022). Cybersyn, big data, variety engineering and governance. AI & Society 37(3): 1163-1177 (published online 4 January 2022)checkedopen access
A first-person account of Cybersyn, written fifty years on and open access. It treats the viable system model, not the technology, as Cybersyn's main contribution; uses Ashby's law of requisite variety to read complexity as variety that an organisation must attenuate or amplify; and describes the author's own move from the VIPLAN method to second-order cybernetics. It lists ten governance questions (for example when to centralise or decentralise, who the stakeholders are and how to include them, how boundaries adapt to distributed demands) and five relationships (performance, cohesion, policy making, inclusion, citizenship). Those questions are a ready checklist for reading a mandate for ability to act. It says governance is needed at all recursion levels from local enterprises to governments, but gives no detailed nesting model, and it does not mention algedonic signals. Read it beside Medina's history (thinking-05): one account from a participant, one from a historian.
Read first: Section 5 ('Reflexion about the VSM and complexity management'), where the ten governance questions are set out, and section 3 ('Real-time management and variety engineering').
- Edgar Morin (2008). On Complexity. Hampton Press, Cresskill, NJ (series 'Advances in Systems Theory, Complexity, and the Human Sciences')checked
For Morin, complex thought rests on three principles: dialogic (opposed terms held together), recursive (products feed back on their producers) and hologrammatic (the whole is in the part and the part in the whole), according to the secondary sources seen. The last matters for the name. Holagram sounds like it, but rule 2 says no level sums up another, while Morin's principle puts the organisation of the whole inside each part. The HOLON page should say which it means. The dialogic principle is nearer to rule 3 and to 'disagreement is data'.
- Werner Ulrich (2022 (book: 1983)). Assessing assumptions about boundaries with critical systems heuristics (i2Insights); source book: Critical Heuristics of Social Planning. i2Insights, posted 24 May 2022; book: Paul Haupt, 1983 (a Wiley edition is listed in Open Library)checkedmust readopen access
The nearest existing practice to the GRAM. Boundary critique asks which judgments define what counts as relevant: who is the beneficiary, who decides, what lies outside the decider's control, whose expertise counts, what counts as success, who speaks for those affected but not involved. Each question is asked as 'is' and as 'ought', and a mismatch marks an unresolved boundary issue. It is meant for non-experts to put to experts, and the page says it is not a way to find right boundaries: it makes assumptions open to challenge. For Holagram: (a) it states 'write it so that someone else can contest it' in a form practitioners know; (b) the is/ought pair gives the mandates method its central contrast, what the text authorises against what the unit would need to act; (c) claiming legitimacy by contestability, not validity, is a safer position than claiming truth. Limit: twelve questions per Entry is too many, and the questions were written for plans and claims, not for reading a document against a place.
Read first: The twelve questions and the passage on 'is' and 'ought' modes.
- Curtis McCord and Christoph Becker (2019). Sidewalk and Toronto: Critical Systems Heuristics and the Smart City. arXiv:1906.02266 (preprint, submitted 5 June 2019)checkedopen access
A preprint that applies Critical Systems Heuristics to an urban planning programme, the Sidewalk Toronto smart city development, asking how its stated values are contested in practice. The abstract reports that powerful stakeholders and beneficiaries limit the project's potential by sidelining residents in deciding its purpose.
- Gerald Midgley and Erik Lindhult (2021). A systems perspective on systemic innovation. Systems Research and Behavioral Science 38(5): 635-670checked
Written for people who run innovation programmes. The abstract sorts 'systemic innovation' into five meanings: technology (innovations combined into a larger product system), policy and governance (shaping policy at local, regional or national scale so that innovation systems have an enabling environment), societal laws and norms, collaborative networks, and a systemic process of engaging people in systemic thinking and action. It argues that the fifth matches most systems thinking of the last fifty years and can improve the other four. For Holagram this is a taxonomy written for people who run innovation programmes. Holagram belongs to the fifth sense, a process for reading and acting systemically, which is a claim about how people read, not about an innovation system, and the page can say so; the mandates method works in the second sense. Systemic Intervention (2000) is the fuller statement of Midgley's boundary critique approach, known to me from the catalogue record only.
- Peter Checkland (1981). Systems Thinking, Systems Practice (soft systems methodology). John Wiley & Sons, Chichester, 1981; reissued 1999correctedmust read
The closest method precedent, and the one a systems-literate reader is likely to name. In soft systems methodology a model of purposeful activity is not a description of the situation; it is a device for debate, built from one declared worldview (the W in CATWOE), set beside the perceived situation, and used to find changes that people can accept. According to the training-site article, Checkland used 'holon' for such models, so the word already carries a practitioner meaning besides Koestler's. Holagram's Reading plus Test is close to model plus comparison. Use it to say in one line what Holagram repeats. 19, as cited by Wikipedia) before the HOLON page says anything about it.
Audit: Say nothing on the HOLON page about Checkland's use of 'holon' until Checkland and Scholes (Soft Systems Methodology in Action, 1990; 1999 ed. p. 19) is read; the cited page ties 'holon' to the actors' worldview, not to activity models.
- Donella H. Meadows (1999). Leverage Points: Places to Intervene in a System. The Sustainability Institute (the page now sits on the Donella Meadows Project site, Academy for Systems Change)checkedopen access
A ranked list of twelve places to intervene, from constants and buffers, through information flows, rules, self-organisation and goals, to the paradigm and, at the top, the power to transcend paradigms. For Holagram it is a second way of stating a level: not a size of unit but a depth of change. It suits the strategy method: each Reading can say where in the list the strategy text intervenes (a parameter, a rule, a goal, a paradigm), and the Test can check whether what the organisation does acts there. The top item, holding any worldview loosely and accepting that none is fully true, is the stance behind 'no reading judges itself'. Practitioners know the list, so Holagram has to say what it adds: it does not rank interventions, it tests readings.
Read first: The whole page (short).
- Horst W. J. Rittel and Melvin M. Webber (1973). Dilemmas in a General Theory of Planning. Policy Sciences 4(2): 155-169checked
The classic planning-theory statement that problems of social policy resist the treatment that suits tame problems. Of the ten characteristics on the Wikipedia page, four bear on Holagram: no stopping rule; no immediate or ultimate test; every problem can be seen as a symptom of another (so the level chosen frames the problem); and the planner has no right to be wrong, being liable for the consequences. They threaten two claims, that a three-week study with a Test can close (held, broke and unchecked may read as closure) and that choosing a level is a technical step. Ackoff's 'messes' (situations made of interacting problems, which managers manage but do not solve) is the management-literature twin of the argument.
Read first: The ten characteristics of wicked problems (as listed on the Wikipedia page).
- Paul Cilliers (2001). Boundaries, Hierarchies and Networks in Complex Systems. International Journal of Innovation Management 5(2): 135-147checkedmust read
A short critique aimed directly at what Holagram draws. Per the abstract: models of complex entities such as organisations are problematic; boundaries of complex systems exist but are unusual in kind; hierarchies matter but are less clearly nested than often assumed; and network models share the structure of complex systems yet carry the limits of all models. A reader's gloss in a search result (secondary) says a boundary should be thought of as something that constitutes what it bounds. Holagram's levels are boundaries and its diagram is both nested and networked, so all three claims apply. Defences: levels are rules about the kind of Trace admitted, not claims about structure; the diagram is a reading device and says it is a model; nothing is aggregated. Held threads on the same ground: holarchy-11 (Potochnik and McGill), holarchy-12 (Wimsatt), holarchy-09 (Tsing).
- Kristiaan P. W. Kok, Anne M. C. Loeber and John Grin (2021). Politics of complexity: Conceptualizing agency, power and powering in the transitional dynamics of complex adaptive systems. Research Policy 50(3): 104183checkedopen access
A critique from inside complexity thinking. The authors try to join socio-material approaches with complex adaptive systems approaches so as to treat the politics of transformation; they define power as relational and productive, give non-humans a part, and set out six mechanisms through which power relations redistribute resources and feed either self-reproducing or transformative system dynamics (abstract, via search summary). Used against Holagram: 'many agents, disagreement is data' does not say whose view can stop an act, and the mandates method is about exactly that. Used for Holagram: it offers a way to write a Reading about ability to act as a statement about power and resources, and to include non-human actors without a separate theory. The article concerns sustainability transitions, not public mandates, so the fit is conceptual.
- C. F. Kurtz and D. J. Snowden (2003). The new dynamics of strategy: Sense-making in a complex and complicated world. IBM Systems Journal 42(3), from p. 462checked
A framework that practitioners in governance and strategy are likely to know, used to say what kind of situation they are in. The abstract says the paper questions three assumptions behind organisational strategy (order, rational choice, intent) and offers Cynefin as a sense-making device from action research on narrative and complexity. It sorts situations into domains; Holagram sorts claims by level; a Reading could carry both labels. What the critics say, as listed on the Wikipedia page and not checked: terms are ambiguous (Williams and Hummelbrunner, 2010), the framework is hard to follow and rests on thin foundations (Firestone and McElroy, first published 2003), and the domain names changed, from known and knowable (2003) to simple and complicated (2007), then obvious (from 2014) and clear (from 2015). The lesson for Holagram is to fix the level names and say how a level is assigned, so that the levels do not meet the same complaint.
- S. V. Subramanian, Kelvyn Jones, Afamia Kaddour and Nancy Krieger (2009). Revisiting Robinson: the perils of individualistic and ecologic fallacy. International Journal of Epidemiology 38(2): 342-360checkedopen access
The authors reanalyse the 1930 US census data that Robinson used for the ecological fallacy, with a multilevel model of individual and state-level variables, and conclude that researchers face both the ecological fallacy (reading groups as individuals) and the individualistic fallacy (reading individuals without their context), so that analysis must be multilevel and grounded in historical and spatial context. For Holagram, a Reading at territory or jurisdiction level cannot be built by adding street or act Entries, and a street Entry is not a sample of the territory: rule 2, stated with evidence. It also says that an Entry about a body or a street still needs its context kept in view. Limit: the setting is statistical analysis of population data; a Holagram study has dozens of Entries and no statistics, so the lesson is about inference, not method.
- Elena Esposito (2022). Artificial Communication: How Algorithms Produce Social Intelligence. MIT Presscheckedopen access
The Luhmann line applied to synthetic agents. The abstract argues that machine learning is better understood as artificial communication than as artificial intelligence: machines add to social intelligence by changing how people communicate with them, not by thinking as people do. If communication is completed by the person who receives and uses it (a reading of Luhmann that the 2026 preprint gives; secondary), then a synthetic hand in Holagram produces text whose status is fixed by the person who accepts it, edits it and logs it. That supports recording the receiver in the hands note, and supports the rule that synthetic agents supply no Traces. The chapters on lists, on reading images and on algorithmic forecasts touch three things Holagram handles: documents as lists, diagrams, and prediction. Limit: Esposito treats algorithms as partners in communication, while Holagram treats them as tools that carry no responsibility; the HOLON page needs to say which of the two it holds.
- Pete Barbrook-Johnson and Alexandra S. Penn (2022). Systems Mapping: How to build and use causal models of systems. Springer International Publishing, Chamcheckedopen access
A baseline for what practitioners can already do when they draw a system: by its title and chapter list the book is a how-to guide. It covers rich pictures, theory of change diagrams, causal loop diagrams, participatory systems mapping, fuzzy cognitive mapping, Bayesian belief networks and system dynamics, with chapters on the data and evidence a map can be built from, on running mapping workshops, and on comparing, choosing and combining methods. Holagram's diagram of linked Entries is a system map of a different kind: the nodes are readings with sources, not variables, and the links record tests, not causes. The comparison chapter gives a way to say that on the page. It is open access, so the audience can check what it says. The contents of the chapters are unchecked.
Read first: 'What Data and Evidence Can You Build System Maps From?' and 'Running Systems Mapping Workshops' (titles only).
- Geert Lovink and Ned Rossiter (2018). Organization after Social Media. Minor Compositionschecked
58) the authors say that arguments over human and non-human agency, of the kind raised by actor-network theory, belong to the first era of cybernetics, which tried to join the human body to the machine and led to decades of argument about artificial intelligence; they ask for a 'cybernetics 2.0' about messy social complexity among people, beyond the mirror dynamic of man and machine. Against Holagram: a dialogue with computational agents may repeat the first-era debate and leave the harder problem, how people organise, untouched. For Holagram: the hands record and the outside reader keep the human-to-human problem inside the method.
What it implies for the proposals
- HOLON page: set the seven lines against their oldest statements and their main limits, and say in one sentence what is repeated and what is added. Level: Bateson [thinking-01], Beer [thinking-04], Meadows [thinking-12]. No roll-up: Cilliers [thinking-14] and Subramanian and colleagues [thinking-17], with Morin [thinking-07] as the opposite view. Many agents: Checkland [thinking-11] and Ulrich [thinking-08]. No reading judges itself: von Foerster [thinking-02] and Luhmann [thinking-03]. Contestability: Ulrich. The added parts are the Trace that someone can open, the hands record, the labelled Fiction, the failure log and the rule against roll-up; say that the rest is inherited. Name the two earlier uses of 'holon' (Koestler's, and Checkland's artificial model, the second from secondary evidence only) and say how Holagram differs from Morin's hologrammatic principle.
- HOLON page, line 5: state a stopping rule for the regress. Second-order thinking [thinking-02, thinking-03] says that the observer of an observer is still an observer, so 'another level or another agent' cannot be shown to be independent. Say that Holagram stops at an outside reader and the log, as a decision, and claim legitimacy by contestability, not validity [thinking-08]. For synthetic agents add the human who received and accepted the output [thinking-18].
- GRAM page: give each Entry a boundary header of about six lines taken from Ulrich's twelve questions [thinking-08]: whose purpose the document serves, who decides, what lies outside the decider's control, whose knowledge counts, who is affected but not involved, what is left out. Ask each in 'is' mode (what the document says) and 'ought' mode (what the reader would need). Show one worked example. McCord and Becker [thinking-09] is a preprint applying it to a planning programme; check its method before copying it.
- GRAM page: put the method beside what the audience already uses, one line each. Cynefin classifies the situation [thinking-16]; leverage points rank places to intervene [thinking-12]; systems mapping builds causal maps and runs workshops [thinking-19]; soft systems methodology builds models from declared worldviews and compares them with the situation [thinking-11]. Holagram's difference to claim: the evidence in each Entry is a source someone can open, each register records its hands, failures are kept, and nothing is rolled up. What it does not do: build a causal model, assign a domain, or make a recommendation.
- Cities (body, street, site, institution, territory). (a) Keep levels as rules about the kind of Trace admitted, as in round one, and add a cross-level inference rule from [thinking-17]: a territory reading is not a sum of street Entries and a street Entry is not a sample of the territory. (b) Use Rittel and Webber's characteristics as a list of failure modes [thinking-13]: no stopping rule (fix the fieldwork block in advance), symptom of another problem (each Reading says why this level and not the next), no right to be wrong (say who could act on the result). (c) Test one planning document by asking Ulrich's 'is' questions of its text and checking the 'ought' answers on the walk [thinking-08, thinking-09]. Domain failure: a document that speaks to every level at once makes the test at another level formal; log it in the first study.
- Strategy making (move, unit, organisation, field, world). Documents: strategy papers, annual reports, KPI sheets, budget books. Ground: what the organisation publishes and does over a stated period. The Test is Beer's rule that the purpose of a system is what it does [thinking-04]: the stated purpose is a Trace about the text, and the purpose is read from outputs. Tag each Entry with a leverage level [thinking-12] to say where the strategy intervenes, and optionally with a Cynefin domain [thinking-16] if the version of the labels is recorded. Failure modes: strategy text written so that no output can contradict it; the field and world levels reachable only through other organisations' documents; Fiction in the house style of a scenario paper, which looks like a source. First study: one public strategy and one year of public outputs for an organisation that is not a client; three weeks; about ten Entries; success if one Reading breaks. See also held documents-15 (Mastop and Faludi, the performance principle for strategic plans).
- Mandates (act, unit, authority, jurisdiction, polity). Read each decree, statute, charter or delegation letter as a decision premise or programme [thinking-03; chapter titles seen: Decision Premises, Decision Programs, Personnel, Uncertainty Absorption]. For each, say which viable-system functions it assigns (operations, coordination, control, intelligence, policy) and at which recursion level [thinking-04, thinking-05, thinking-06]; Espejo's ten governance questions and five relationships [thinking-06] are a ready prompt list. Ability to act is then the 'is'/'ought' gap [thinking-08]. Test at another level: look in the public record of acts (appointments, tenders, regulations, reports) for the function the text omits; a missing function that someone else performs breaks the Reading 'the unit cannot act'. Failure modes: reading boilerplate as intent; documents from one drafter or template, so the levels are not independent; the viable system model read onto the text as five boxes; overlapping jurisdictions, which break the nest (see held holarchy-14, Hooghe and Marks). Midgley and Lindhult's governance sense of systemic innovation [thinking-10] names the setting.
- Power and liability: add to every Test a line saying who could act on the result and what being wrong would cost them [thinking-13], and add to every mandate Reading a line saying who can say no and by what means [thinking-15]. Keep the unit of reading the institution, the act or the document, not a named person. This is the care rule for governance material, and it keeps the method clear of the 'no right to be wrong' problem for individuals.
- Hands: for each human or synthetic hand, record the worldview in which it writes or tests (Checkland's W [thinking-11]) and one line on what its position cannot see [thinking-02]. For a synthetic hand also record who received, edited and accepted the output [thinking-18]. Require that the 'other agent' in a Test differs in provenance from the author (a human outside reader, or a model from a different family), since shared training data makes independence formal (see held agents-10).
- Two first-study options, public material only. (a) Mandates: one public charter or decree and the public record of one unit's acts over a year, ten Entries, success if one function is found elsewhere or one clause proves empty. (b) A Cybersyn desk study: read the viable-system story in Medina [thinking-05] and Espejo [thinking-06] as a set of Entries (a diagram of a state, a promised autonomy, a coup) to rehearse the method on a case where the evidence is already published and nobody involved is a client. Do not use the second as evidence about any present programme.
Ideas to take
Open the HOLON page with a table of the seven lines against their oldest statements and their main critics, and claim only the writing form as new.
Changes: The page pre-empts the remark that this is soft systems methodology and boundary critique under new names. It says what is inherited (levels, observer inclusion, contestable boundaries, models as devices) and what is added (a Trace someone can open, a hands record per register, labelled Fiction, a failure log, no roll-up). It also names the two earlier uses of 'holon' and the hologrammatic principle.
Risk: The table rests on abstracts, contents lists and secondary summaries for most items, and a reader who knows these traditions will check it against the books. Claiming less novelty is safer but weakens a pitch. Checkland's use of 'holon' is from secondary sources only.
Changes: Rule 5 gets a form that can be checked in the log, borrowed from Bateson's logical types. The failure log gains a class, same-level test, beside the existing ones, and the first study can count how often it occurs.
Risk: Logical types concern messages, not places or institutions, so the check is an analogy. Cilliers and Subramanian say levels are not neatly nested, so two levels can share one cause or one drafter and the check becomes formal. Bateson himself was reported as doubting that the top level of learning could be tested.
Put a short boundary header on every Entry, adapted from Ulrich's twelve questions in 'is' and 'ought' mode, and use the gap between the two as the finding for mandates.
Changes: The GRAM gets a contest route that practitioners already know: whose purpose the document serves, who decides, what lies outside the decider's control, whose knowledge counts, who is affected but not involved. In the mandates method the 'is' answer is what the text authorises and the 'ought' answer is what the unit would need to take the next step, so ability to act is the gap.
Risk: Twelve questions per Entry is too many for a three-week study, and six is a guess. The questions were written for plans and claims, not for reading a document against a place. The page that states them says boundary critique does not find right boundaries, so Holagram must not claim validity either.
Read a mandate as a set of decision premises, say which viable-system functions it assigns and at which recursion level, and test the reading against public outputs with Beer's rule that purpose is what a system does.
Changes: Gives the mandates levels (act, unit, authority, jurisdiction, polity) a concrete reading frame and a Test that uses public material only: find where the omitted function is performed, and compare the stated purpose with what the unit publishes or does. Espejo's governance questions become the prompt list.
Risk: Any organisation can be read as five boxes, which is the self-confirming reading the method is built to prevent; the Test has to look for the function being performed, not for the box. Maturana objected to extending autopoiesis to society, and the same objection can be made to 'viable system' for a ministry.
Add two lines to every hands note: the worldview in which the hand writes or tests, and what its position cannot see; for a synthetic hand add who received and accepted the output.
Changes: The hands record becomes a short statement of the observer's position, which is what second-order thinking asks for, using Checkland's declared worldview. For synthetic agents the receiver is recorded because, on the Luhmann line, meaning is completed by the one who receives the text.
Risk: 'What its position cannot see' is itself a Reading by the author about the agent. Esposito treats algorithms as partners in communication and Holagram treats them as tools without responsibility; the page must say which it holds.
State the no-roll-up rule as a two-way fallacy rule, and run one deliberate roll-up in the first study to see what breaks.
Changes: Any sentence that moves between levels says which error it risks: reading a group as an individual, or an individual without the group's context. The stress test shows whether the rule can be kept in the write-up. Morin's principle that the whole is inscribed in the part is the opposite view, so the HOLON page says where Holagram differs, which also deals with the likeness of the name to 'hologrammatic'.
Risk: The higher levels will stay thin and the write-up must generalise somewhere. The fallacy literature concerns population statistics, and a Holagram study has dozens of Entries, not data sets.
Use the Cybersyn story as a standing example of a government diagram, and rehearse the method on it as a desk study before any fieldwork.
Changes: Gives the HOLON and mandates pages a published case in which a diagram of an institution had political reach, was argued over and was ended by a coup, and a rehearsal in which every Trace is already public. Reading Medina's history beside Espejo's first-person account is itself a test across two hands.
Risk: The history is politically loaded and some readers will hear an endorsement of cybernetic government. One case proves little.
Changes: Gives the page a way of stating a hard rule without a lecture, a precedent for the Fiction register from the systems tradition itself, and a signal that the site is a research style and not a consultancy.
Risk: It can read as pastiche, and Bateson's metalogues are of their time. Keep it to one short piece on one page and label it as fiction.
Counter-arguments
- This is soft systems methodology and boundary critique renamed. Checkland's models are devices for debate tied to a declared worldview and compared with the situation [thinking-11]; Ulrich's boundary questions let those affected contest the boundaries [thinking-08]; Midgley and Lindhult define the systemic sense of innovation as a process of engaging people in systemic thinking [thinking-10]. A systems-literate practitioner will ask what is left once these are removed. The honest reply is the narrower list: the Trace someone can open, the hands record, labelled Fiction, the failure log and no roll-up.
- Boundaries and hierarchies are not given. Cilliers argues that boundaries of complex systems exist but are unusual in kind, that hierarchies are less clearly nested than assumed, and that network models share the limits of all models [thinking-14]; a reader's gloss says a boundary constitutes what it bounds. Holagram's five levels are boundaries and its diagram is both nested and networked. If the levels are read as a claim about reality, all three points apply. Held threads on the same ground: Potochnik and McGill (holarchy-11), Wimsatt (holarchy-12), Tsing (holarchy-09).
- Wicked problems have no stopping rule and no right to be wrong [thinking-13]. A three-week study ending in held, broke or unchecked can look like closure on a question that has none, and a note that reads a mandate as unable to act may be used by someone who then bears the consequences. Each wicked problem is also a symptom of another, so the choice of level is part of the problem, not a technical step.
- The regress. Second-order cybernetics and Luhmann say an observer of an observer is still an observer [thinking-02, thinking-03]. 'Another level or another agent' does not leave the frame: two levels can share one drafter, one template or one cause, and two language models share training data. If meaning is completed on the receiver's side [thinking-18], the chain of tests ends in one human head.
- Complexity thinking lacks a theory of power. Kok, Loeber and Grin make the point from inside the field [thinking-15]. 'Many agents, disagreement is data' treats positions as comparable, and boundary critique claims legitimacy, not validity [thinking-08]. A mandates method is about who can say no, and Holagram as written has no way to say it. Scott (documents-13, held) adds that making a society legible is itself a way of seeing from above, and a diagram of linked Entries is a legibility device.
- Borrowed terms. 'Holon', 'recursion' and 'viable system' applied to a ministry or a free zone can draw the same objection: biological and engineering terms work as metaphors that borrow authority.
- Level schemes and situation frameworks that cannot be tested. The critics listed on Wikipedia's Cynefin page (not opened) say its terms are ambiguous, that it is hard to follow and that the domain names changed between 2003 and 2015 [thinking-16]. Holagram has three level schemes (body to territory, move to world, act to polity) with different names; a critic can call them three ad hoc ladders. The method needs one rule for how a level is assigned and how that assignment can fail.
- The Entry looks like a stepwise process. Lewis, McGann and Blomkamp (Policy and Politics 48(1), 2020; Crossref abstract seen; not listed) report that design thinking in public sector innovation labs meets obstacles when it meets power and politics. The answer has to be that the Entry is a writing form and a log, not a process model, and the page should say so.
- Human and machine partnership is an old debate. A reader of this view will say Holagram's dialogue with synthetic agents spends effort on the wrong pair.
- Practitioners want something to do on Monday. Rittel and Webber's wicked problems and Ackoff's messes both leave managers with situations to manage, not problems to solve [thinking-13]. A log of readings and failures does not tell a strategy unit or a lab what to change. The method needs a stated use: it improves what is read before action, it shows which reading broke, and it makes no recommendation.
Gaps
- Cynefin critics: two web searches and two Crossref title searches found no peer-reviewed critique. The record holds applications and two papers by French in the Journal of the Operational Research Society (64(4), 2013; 66(10), 2015; records and abstracts only) that use the framework for decision analysis and scenarios. The critiques named on Wikipedia's Cynefin page (Firestone and McElroy; Williams and Hummelbrunner, whose 'Cynefin' chapter in Systems Concepts in Action has a Crossref record) were not opened. A 2026 article in Medical Humanities with 'Cynefin' in its title (Downs) is about the Welsh word, not the framework.
- Checkland's 'holon': only secondary evidence (Wikipedia's page on soft systems methodology, which cites Checkland and Scholes 1999, p. 19; an unsigned training-site article dated 7 September 2026; SEBoK). A link to Koestler was not found. Check the books before the HOLON page says anything about it.
- Morin: no text seen; the three principles come from a wiki page and a thesis in search results; the wording that the whole is also less than the sum of its parts was not found. No Italian edition of On Complexity confirmed.
- Reidel, 1980) and The Tree of Knowledge (Shambhala, 1987); Churchman, The Systems Approach and Its Enemies (Basic Books, 1979); Midgley, Systemic Intervention (Kluwer/Plenum, 2000); Beer, Designing Freedom (1974) and Brain of the Firm (1972). Jackson, Critical Systems Thinking and the Management of Complexity (Wiley, 2019) has an Open Library record, but Wiley's page returned 404, so its typology of problem contexts is unchecked.
- Habermas and Luhmann, Theorie der Gesellschaft oder Sozialtechnologie: Was leistet die Systemforschung? (Suhrkamp, 1971; Open Library record seen): the classic European critique of systems theory as social technology.
- Ulrich, Critical Heuristics of Social Planning (1983): catalogue record only. Check the exact wording before quoting.
- Most had no hits. No text by Bateson, von Foerster, Luhmann, Beer or Ulrich. The Churchman hit in Cities by Design is a different Churchman, writing on perceived density.
- No news or programme claims are made. Nothing was checked about the UAE, Dubai, Singapore, Estonia, the UK or the EU, and no government programme is described; the Cybersyn items are history and a first-person account.
- Language: the Italian routes verified at catalogue level are Bateson (Verso un'ecologia della mente, Adelphi) and La sfida della complessità (Atlan, Bocchi, Ceruti; Feltrinelli, 1985, 435 pages; contents not seen). Von Foerster's Italian edition is recalled only. No Finnish item. Estonian: leads only (see cross-leads).
- Search effort: about 20 web searches and about 60 page or record fetches (Crossref, Open Library, publisher, archive and wiki pages). The ones that failed are listed under access limits.
- Access limits: Springer's book page redirected to a login; the MIT Press, Chatham House, OECD and University of Hull pages returned 403; the New Yorker is blocked in this tool; Wiley's page returned 404. For these, verification rests on Crossref, Open Library and search summaries.
- Open items for the main session (not recorded by me): decide how the HOLON page names Checkland's and Koestler's 'holon' and Morin's hologrammatic principle; decide how many boundary questions an Entry header carries; read Rittel and Webber, Cilliers and Ulrich's primer in full before they are cited on the site; check Checkland and Scholes p. 19; check the full text of Kok and colleagues before using 'powering'; decide whether the Habermas and Luhmann exchange is worth a verified entry.