(2006-06-07 modification of submission for HIC Business 2006)
Greald Henstra
Groningen University, faculty of Management and Organisation
For many people collaboration is equal to making use of each others knowledge. In many cases it is not necessary to transfer knowledge itself; a simple derivative instruction usually is sufficient. Organisations then rely on their people who just deploy their knowledge whilst keeping its essence to themselves. Indeed, in reproductive processes task repetition usually provides organisations with enough opportunities to have knowledge disseminated in ways like Nonaka's SECI cycle[1].
In processes like in innovation projects however, very sparsely performed tasks may be invoked. In those cases, it is widely recognised to be a major problem to localise and address suitable knowledge.
In innovation projects knowledge often is exploited to deal with 'problem solving'. Conversely, 'problem solving' processes also have proven to be useful to explore new knowledge. They thus may hold the key to elucidate ways to trace appropriate knowledge in organisations.
A vast volume of literature apparently agrees on the nature of 'problem solving' or 'designing': it is considered as the process by which designers search for sets of means, solutions to the 'problems', to attain a certain set of specifications or purposes.
A design, in any stage of development, consequently can be represented by (C:) a set of purposes, (F:) a set of problems to be solved, and (P:) a set of solutions to the problems or, equivalently, a set of means that effectuates the purposes.
In practice the separation between the three is not obvious however. During the development of a design its representations grow interdependently and continually transform into one another.
But, in spite of their entanglement, the design representations may be helpful for analysing problem solving processes and the knowledge associated with them. For, if any state of a design can be indicated by its representations (C, F & P), any design transformation can be so by the representations of both its input and output states.
This has given rise to the set up of a framework connecting networks of design processes and knowledge carrying people. This framework is helpful in both their traceability and analysis of these networks.
In processes like in innovation projects
"usually knowledge is gained at an operational level, but generally it is not systematically recorded in the firm's corporate information systems. It may be disregarded, for instance because it's not directly oriented towards the enforcement of the market position[2]. On the other hand, being newly acquired knowledge, it may not fit in the data base structures at hand. It remains tacit. [..] Knowledge remains 'tacit' if it's not [codified, so certainly it remains 'tacit' if it's not] communicated: codified, transmitted, received and decoded[3]."[4]
Our problem is to localise knowledge, which boils down to two conditions.
One condition is assumed to be to monitor the knowledge creation process, because its traces can be expected to be left in the monitoring structure.
The other condition considers retrieving the once observed knowledge. The monitoring system should have 'two way accessibility'; knowledge localised or reported in the structure should be easily retraced and decoded.
The 'codes' of both sender and receiver should coincide; which is not self-evident, especially in the hazy starting phases of a new development.
Both conditions induce the need for a common code, as was explained by Rapoport[5]: "[*=]The 'substrate' of mathematical information theory is a population of symbols and probability laws governing the frequencies of occurrence of messages conceived as strings of these symbols. [=*]
In social communication, content is of prime importance. The problem of estimating frequency distributions of messages is, accordingly much more difficult. For, to estimate these probabilities, we must know the entire repertoire of messages. In the context of technical communication theory this repertoire is known, because it consists merely of strings of alphabet units. In communication with semantic content, we are dealing with a repertoire of ideas, which is nowhere given in terms of strings of units.
To illustrate this crucial difference, consider the game of Twenty Questions[:] guessing what someone had in mind by getting answers to yes/no questions. [..]
Now if the guessing could proceed letter by letter, then any finite message could be guessed after a finite number of questions. [But otherwise] this problem [..] involves a matching of ideas, experiences, feelings, and so on. [P]robabilities [then] cannot be estimated in any straightforward manner, since probabilities involve not merely what was sent but what could have been sent, hence the repertoires of the categories that are only vaguely known. [..]
[*] The substrate of a [..] theory of human communication [-analogous to mathematical information theory-] ought to be a repertoire of 'ideas' and the probability laws governing their association.
Neglect of this fundamental difference can lead astray attempts to utilize information theory in the social sciences."
Can a ~common code~ be found to overcome these constraints?
In innovation projects knowledge often is exploited to deal with 'problem solving'[6]. Conversely, 'problem solving' processes also have proven to be useful to explore new knowledge. They thus may hold the key to elucidate ways to trace appropriate knowledge in organisations.
Since a long time a vast volume of literature[7] is apparently in agreement on the nature of 'problem solving' or 'designing': it is considered as the process by which designers search for sets of means, solutions to the 'problems', to attain a certain set of specifications or purposes.
Yet its scientific value is not established firmly. "Design methods are [..] the result of collective experience and understanding. They are based on [..] descriptive analyses of design and development processes [and] on [..] logical reasoning. Although they are not [(entirely)] built on authority, tradition or intuition - three subjective ways of reasoning that are hard to criticise rationally [. But] systematic evaluation of the effectiveness of design and development methods so far has hardly been undertaken."[8]
One of the consequences is that there is relatively little knowledge about the field of design and development. Alternative ways of working, among which practices in the field[9], are virtually unknown.
Indeed "our selection of the variables to manipulate in any given situation is conditioned by our previous experience and our education. As a result we [humans] unconsciously preclude from considering certain types of controllable variables and confine ourselves to certain traditional ones"[10].
An argument goes that "one of the reasons [why design methods are not evaluated systematically] is that product development is a historical process, which in practice cannot be repeated."8 Apart from the fact that it is very hard to perform any design method in practice in its pure intended form "this makes it hard to prove that alternative methods would have yielded a better result."8
In other words: all development projects are unique, or, alternatively: they lack a clear common frame of reference.
The argument may be prompted by the widely adopted attitude[11] in organising development-processes for effectiveness and efficiency. As reflected in the tendency to optimise the "mortality curve"[12] of new-product ideas in selecting ~good~ ideas from ~bad~ ones as soon as possible, this attitude could be recognised as the repression of "type II errors".
But as usual this goes at the cost of "type I errors", which apparently is neglected.
In case of abundance of ideas a waste as such may be affordable. Otherwise[13] an alternative attitude should be considered. "Let a hundred flowers bloom, and let the course of events decide whether they are fragrant flowers or deadly weeds."[14]
Nevertheless, innovation involves pioneering in unknown territory which is a risky endeavour. So, to search for more refined routines a new orientation towards practice would be helpful. This, again, calls for a common frame of reference.
Thus a common frame of reference, one of the elements of the 'common code' we are looking for, might alleviate comparison of development projects and design methods, in practice as well as in theory. For that reason in the following section an attempt is made to set up such a code.
Consider a 'design'[15] , its 'designer'[16], and the 'problem owner'[17].
A1: Let EΩ be the set[18] of all 'effects', 'cosmonomically'[19] possible phenomena that are dependent from other 'cosmonomic' phenomena.
Let SΩ be the set of phenomena that are conditions for elements of EΩ.
Then there must be a set of 'relations' RΩ in such a way that
for all e∈EΩ there is a set S⊂SΩ for which:
R:S→e
or, alternatively written,
R(S)=e
([20])
(As a matter of fact this is a model that represents everything in existence.)
For any subset Ei⊂EΩ there consequently must be a Pj⊂SΩ with Fk⊂RΩ in such a way that:
Fk:Pj→Ei
or, written alternatively
Fk(Pj)=Ei
(19), (i,j,k∈ℕ)[21]
"According to the standard model for analysing human activity, predominant in philosophy since David Hume, and basis for rational decision theory, any human action can be explained with the assumption that the actor has a certain desire and certain beliefs.
A2: Whenever the actor has the desire c, and the belief gf (34) that action p will result in c, (s)he will perform p." [22]
Let C be effects desired by 'problem owner'16 = 'desired effects' = 'desires', and CRe=C∩EΩ, the desires that principally can be fulfilled.
A3: If subsets Ei can be found containing CRe they usually also contain 'side effects' Ni in such a way that:
Ei=CRe∪Ni
([23])
'Side effects' may be 'not desired' explicitly but also can be 'undesired': NOi⊂Ni as ~every particular solution Ei has its constraints~.
A complete set CRe is assumed to contain the negation or neutralisation of all NOi
A set of goals, objectives, 'desired effects' CRe⊂Ei, either or not explicitly recorded in (a list of) 'criteria' has been realised if a set of means or solutions Pj⊂SΩ is produced in such a way that by certain 'functions'[24], transformations[25] or processes Fk⊂RΩ are deployed in such a way that:
Fk:Pj→CRe
or, written alternatively,
Fk(Pj)=CRe
(19), (i,j,k∈ℕ)20
For any CRe at least one Ei exists, and consequently at least one set of means Pj and likewise at least one function set Fk.
A4: But the probable numbers for Ei, Pj and Fk are not necessarily confined to one. Usually a vast number of alternative sets can be distinguished[26], giving rise to a set of sets E⊂EΩ:
E={Ei for which CRe⊂Ei}
and
('preconditions') P⊂SΩ with
P={Pj whith Fk such that Fk(Pj)=Ei=CRe∪Ni}
and
'functions' F⊂RΩ with
F={Fk with Pj such that Fk(Pj)=Ei=CRe∪Ni}
To a 'designer', C represents the goals to be achieved. Unlike C, F is undetermined a priori; the fulfilment of F therefore represents the design problem[27]. As soon as P and C are determined so is F. Thus P will be seen as the solution to F.
A 'design', in any stage of development, consequently can be represented by (C) a set of purposes, (F) a set of problems to be solved, and (P) a set of solutions to the problems or, equivalently, a set of means that effectuates the purposes.
This assertion will be referred to as the 'FPC-model'.
Summarising, assuming A1..A4 the FPC-model states that F,P and C are necessary classes to fully describe a 'design'.
'Realising' a solution Pj means:
selecting[28] one alternative Pj from P from SΩ,
and 'realising' function Fk means:
selecting27 one alternative Fk from F from RΩ
in such a way that
CRe⊂Ei is met "satisfycingly"[29].
A5: Many s∈SΩ depend from something else: SΩ and EΩ are not disjoint. In practice it's safe to assume SΩ=EΩ.
In that case Pj∈EΩ and to have 'realised' CRe⊆k(Pj) first Pj⊆kl(Pjm) must have been 'realised'.
l,m∈ℕ(20)
Fk(Pj)=CRe∪Ni
then becomes
Fk(Fkl(Pjm))=CRe∪Ni.
Reasoning likewise for the new combination and for its successors leads to:
Fk(Pj) = Fk(..(F(k..)(P(j..))..) = CRe∪Ni
proving that 'designs'14 in fact can be represented as hierarchical structures, 'means-end-hierarchies', and consequently can be "decomposed"28.
F, P and C were found to be necessary to fully describe a design. But are they sufficient as well to do so?
A..: Clearly a set Ei is sufficient indeed, by its definition.
Then, since Ei=CRe∪Ni it follows that
CRe is not sufficient on its own.
But also
Ei=Fk(..(F(k..)(P(j..))..)=Fk(Pj)
implying that
P and F combined are sufficient to fully describe a design.
"The creation of complex objects, whether houses or horses, demands two kinds of specification: one for the components and one for the system that guides their assembly"[30].
"A [..] product can be understood as both the product as a whole − the system − and the product in its parts − the components. This distinction has a long history in the design literature[31]."[32]
A 'recipe' is a prescription to obtain a 'desired effect' C, and
(A6:) consists of a list of 'constituents' and an 'architecture': a prescription how to order these 'constituents'.
'Constituents' apparently are represented in P, and 'architecture' in F as it establishes the relations for P to attain C.
A7: In practice the separation between F, P and C is not always obvious. During the development of a design its representations grow interdependently and continually transform into one another. In moving through means-ends-hierarchies means end ends change places: what is an end in one link in the chain is a mean in the next.
PF boundaries
E.g. 'constituents' ∈Pj obviously are a necessary condition to have 'functions' ∈Fk performed. But their mere presence is not sufficient. (E.g. a Volkswagen Beetle is not obtained just by bringing its parts together, not even by shaking them firmly.) The 'constituents' can be transformed into a 'desired result' only if they are ordered correctly. The set of 'constituents' completed with the rules of the 'architecture' is necessary to have the intended 'functions' Fk performed.
Thus both 'constituents' and 'architecture' can be classified as a part of Pj: conditions for CRe=Fk(Pj).
P as F
E.g. ordering takes place to enable the performance of functions F⊂RΩ. The 'architectural' rules follow directly from the functional description. Their close connection also argues to classify the 'architecture' in F.
F as P
E.g. also in case the sets of 'constituents' completed with the rules of the 'architecture' are sufficient to establish Pj, i.e. that transformation Fk necessarily will take place, i.e. Fk is determined unambiguously. Then the separation between Fk and Pj becomes troubled.
PC boundaries
P as C
E.g. in case, descending a means-ends-hierarchy, a Fkl(Pjm) is searched in such a way that Fkl(Pjm)= Pj
Pj then is perceived as a 'desired result'∈C.
C as P
E.g. if there is no stimulus to search for new solutions, ready available solutions or commodities may play the role of a particular desire. E.g. "technical specifications" come to meet this need.
FC boundaries
C as F
E.g. as C can be confused with P and P with F so can C with F.
F as C
E.g. if there is no ready available solution at hand a desired effect may be described in terms of a process to produce it. This is often referred to as a "black-box-approach" or "functional specification".
Even though in particular cases boundaries cannot be decided upon unambiguously, a design in progress still can be recorded in the FPC-model. After all, the model states that a design is represented in all classes F, P and C simultaneously, independent from each other. In case of doubt between classes simply all of them can be considered to apply.
A8: In order to select Pj from P and its connected Fk from F first they need to be found and to be found they usually need to be searched[33] for.
A9: The search for Pj∈SΩ and Fk∈RΩ to fulfil C∩EΩ usually takes place in the "domain of the 'designer's'15 mind"[34].
In this process the 'designer' cannot draw from EΩ, SΩ or RΩ directly but only from what he believes or thinks to know: the sets of 'assumptions' on E, S, R.
assumptions and justifications
A10: These assumptions may be 'certain' or not ([35]).
'Assumptions' γE, γS, γR may or may not be ⊂EΩ, SΩ, RΩ respectively.
Conversely 'cosmonomic elements' of EΩ, SΩ, RΩ may or may not be 'assumed'.
If so
γE∩EΩ, γS∩SΩ and γR∩RΩ
are 'justifyable',
and
'certain' if proven in confrontation with "cosmonomy":
κE, κS, κR ([36])
where
κE⊂(γE∩EΩ),
κS⊂(γS∩SΩ),
κR⊂(γR∩RΩ).
With γE'∩EΩ, γS'∩SΩ en γR'∩RΩ ([37]) is referred to what neither is known nor surmised.
'Delusions' on the other hand in γE∩E'Ω, γS∩S'Ω and γR∩R'Ω (36) exist in the mind but do not in reality.
Only 'certain assumptions' (κA) can be labelled correctly.
The 'justifyable' but not (yet) 'certain assumptions' (γA) cannot be distinguished from 'delusions'.
In spite of their apparent entanglement, the 'design representations' (F, P & C) may be helpful for analysing problem solving processes and the knowledge associated with them. For, if any state of a design can be indicated by its representations (F, P & C), any 'design transformation' can be so by the representations of both its input and output states.
The state of a 'design' at any point t in its development would be represented by its states:
F(t),
P(t),
C(t),
and at some subsequent point t+Δt:
F(t+Δt),
P(t+Δt),
C(t+Δt).
Any 'design transformation' then could be represented by a vector (ΔF,ΔP,ΔC) with:
ΔF=F(t+Δt)-F(t),
ΔP=P(t+Δt)-P(t),
ΔC=C(t+Δt)-C(t)
([38]).
A11: The changes Δ may vary in 'significance' as well as in their 'certainties'.
Furthermore the direction in the means-end-hierarchy provides freedom to the Δ's.
framework
Let for instance 'design transformations' be expressed in a vector with components ΔF, ΔP, ΔC
and let
significance ∈{true, false}
and
certainty ∈{∅, γ, κ}×{γ, κ} ([39]).
Then every vector component can assume 7 values.
In this proposal, a space is set where, as the components are taken to be independent, 73=343 'design transformations' can be distinguished.
illustration
E.g. take "the basic design cycle"[40]: analysis, synthesis, simulation, evaluation and selection.
"Analysis" is the activity where the function is transformed into criteria. It may take the form:
(γ,x2,x3)→(y1,y2,γ), where x2,3,y1,2∈{∅,γ, κ}
"Synthesis" is the activity where a solution is proposed. It may take the form:
(γ,x2,γ)→(y1,γ,y3), where x1,2,y1,3∈{∅,γ, κ}
"Simulation" starts with a suggested solution and is aimed at predicting its effects. It may take the form:
(x1,γ,x3)→(y,γ,γ), where x1,3,y∈{∅,γ, κ}
"Evaluation and selection" involve the comparison of the design's effects and its desired effects. It results in a decision with high certainty and may take the form:
(x,γ,γ)→(y,γ,κ), where x,y∈{∅,γ,κ}
project monitoring
This has given rise to the set up of a framework connecting networks of design processes and knowledge carrying media - usually people.
Empirical data on ΔF, ΔP, ΔC could be acquired by descending "means-end-hierarchies" in the field.
Usually developments are carried out by project-members within projects. By questioning them on their sources (usually colleagues) of knowledge, new to them (γF,γP,γC), on subsequent subjects three interwoven networks can be recorded:
- a network of collaborating people and their knowledge,
- a network of subjects (in the means-end-hierarchy) within the design in progress,
- a network of 'design transformations'.
Networks of people and their knowledge on design subjects, as well as the ones of 'subjects' alone, obviously are immediately useful for operation and management[41].
After all the networks themselves provide 'two way accessibility' for a 'common code' required for communication within the project at hand, just like directory structures do in ordinary file management systems.
Networks of 'design transformations' may offer 'common codes' between projects and sub-projects. Subsequent vectors in the 'design transformation space' linked together can be expected to form graph-theoretical paths[42] ready for analysis[43].
Complemented with data on dependent variables, e.g. project performance[44], hypotheses e.g. on effectiveness and efficiency can both be generated and be tested.
"citation"
[citation interruption]
'defined term or self-citation'
definition
A# assumption
conclusion
~to be refined~
[2] Coombs & al. (1998) for instance found that for certain cases "[] one of the difficulties in 'knowledge management' is that teams of people are assembled to do the work, and they acquire a great deal of experience and skill which appears to be very specific to 'getting that project done' but which in fact, conceals some potentially generic and transferable lessons. But the culture of such R&D work is that the completion of the hardware is the driving force, and after that is complete, the teams are broken up and there is low motivation to reflect on and document the transferable experience and re-use it in future projects. This leads to frequent instances of reinvented wheels."
[5] Rapoport, 1986, pp.142-144
[6] e.g. Boersma, 2002, Coombs, 1998
[7] e.g.: Ulrich, 2000; Roozenburg, 1998; Dorst, 1997; Pahl & Beitz, 1996; Simon, 1996, Eekels, 1995; van der Zwaan & van Engelen, 1994; Suh, 1990; Schön, 1987; Yoshikawa, 1987; Eekels, 1982; Ackoff, 1978; Keeney, 1976; Neumann, 1947
[8] Roozenburg, 1998 (translated from Dutch, GH)
[9] (best practices as well as fair, bad and worst ones)
[11] e.g. Wheelwright &Clark, 1992, Cooper, 1985
[12] Booz-Allen and Hamilton, 1968
[13] see for instance Miller & Morris, 1999
[14] Lau Nai-keung explaining Deng Xiaoping in China Daily, 06-06-06, p4.
[15] 'design' = artefact, in "domain of the mind" or in "cosmonomy", that solves a problem to realise a goal or purpose.
[16] 'designer' = design creating medium
[17] 'problem owner' = person with a need that the design is to fulfill. Usually a customer.
[18] In order to keep the reasoning sharp it's expressed in set theoretical code.
[19] 'cosmonomy'= "the invariant pattern of transformations in nature" (Roozenburg, 1998, p435) or "the autonomous occurrence in nature" (Roozenburg, 1998, p38). Translated from Dutch.
[20] r:x→y or r(x)=y: the relation r between x en y implying that if x exists then y is expected.
[22] adapted from Philipse, 2004, p.3, translated from Dutch
[23] "T1 + action => T2 = effect + side effect", T1,2 respectively are start state and end state; Roozenburg (1998) p.68 (translated from Dutch, GH)
[24] Roozenburg (1998, p62: "a product's technical function is the consciously and intentionally connected power to transform the product's environment" (translated from Dutch, GH)
[25] "transformation" = 'set of "transitions" ' (Ashby, 1956, §2/2)
[26] F={Fk} : "problem space" ; P={Pj}: "solution space" (Simon, 1996)
[27] e.g. Pahl & Beitz, 1996, p31
[31] Alexander, 1964; Marples, 1961
[32] Henderson (1996) p.360-361
[33] even in case of "serendipity" the mind should be receptive
[35] γA = the 'assumptions' on A with indefinite 'certainty' in the "domain of the mind".
[36] κA⊂γA certain 'assumptions' on A in "cosmonomy"
[38] In this case is dealt with variables on a non-parametric scale. Any arrhythmic operation like subtraction should be treated metaphorically.
[39] g symbolises the state of 'assumption', either or not certain; k symbolises the state of 'certain assumption' (="justified true belief" = knowledge, acc Plato). A falsification results in the highest certainty level. ∅ stands for no idea.
[44] e.g. Griffin, 1996; Cooper, 1985; Hollander, 2002
Ackoff, R.L., 1978, The art of problem solving: accompanied by Ackoff's fables, Wiley-Interscience
Alexander, C., 1964, Notes on the synthesis of form, Cambridge MA: Harvard University Press. In: Henderson (1996)
Ashby, Ross W., 1956, An introduction in cybernetics, Chapman & Hall ltd., London
Boersma, S.K.T., 2002, Management van kennis, Koninklijke van Gorcum, Assen
Booz-Allen and Hamilton Inc., 1968, Management of new products, Chicago
Carré,B., 1979, Graphs and networks, Oxford applied mathematics and computing science series, Clarendon Press, Oxford, ISBN 0.19.859622.7
Cohen (Michael D.), Lee S. Sproul, 1996, Organizational learning, SAGE, Thousand Oaks CA, ISBN 0.8039.7088.9
Cooper, R.G., 1985, Selecting winning new product projects: Using the NewProd system, Journal of Product Innovation Management, Volume 2, Issue 1, March, Pages 34-44
Coombs, R., Hull, R., 1998, Knowledge management practices' and path dependency in innovation, Research Policy 27
Cross, N., 2000, Engineering design methods: strategies for product design
John Wiley & Sons
Dorst, K., Describing design, PhD-thesis, Delft, ISBN 90.9010822.X
Eekels J., 1982, Industriële doelontwikkeling, Delftse Universitaire Pers, Delft
Eekels.J, W.A.Poelman, 1995, Industriële doelontwikkeling, Lemma, Utrecht, ISBN 90.5189.435.X
Fauconnier, G., 1986, Algemene communicatietheorie: een overzicht van de wetenschappelijke theorieën over communicatie, Martinus Nijhoff Wetenschappelijke en Educatieve Uitgevers, Leiden
Griffin,A and A.L. Page, 1996, PDMA success measurement project: Recommended measures for product development success and failure, Journal of Product Innovation Management, Volume 13, Issue 6 , Pages 478-496
Henderson R. M., 1990, Architectural innovation: the reconfiguration of existing product technologies and the failure of established firms. Administrative Science Quarterly, 35, 9-30
Henderson (Rebecca M.), 1996, Technological change and the management of architectural knowledge, in: Cohen (1996) pp.359-375
Henstra, G., J.M.L. van Engelen, 1999, The QFD structure for mapping knowledge, transactions from the 11th Symposium on Quality Function Deployment, Novi, Michigan, ISBN 1889477117
Hollander,J, 2002, Improving performance in business development-Genesis a tool for product development teams, PhD thesis, Groningen. ISBN 90.75092.35.0
Keeney, R.L., H.Raiffa, 1976, Decisions with multiple objectives, Operations Research 22, pp.22-24, in: Ormala,1986
Kratzer,J, R.Th.A.J.Leenders and J.M.L.van Engelen, 2004, Stimulating the Potential: Creative Performance and Communication in Innovation Teams, Creativity and Innovation Management, Vol.13,Issue 1,P.63
Marples, D.L., 1961, The decisions of engineering design. IEEE transactions on Engineering Management, 8, 55-71. In Henderson (1996)
Mattick (John S.), 2004, The hidden genetic program of complex organisms, Scientific American, October, pp 30-37
Miller, W.L., L.Morris, Fourth generation R&D, John Wiley & sons, New York, ISBN 0471240931
Neumann, L.J. von, Oskar Morgenstern, 1947, Theory of games and economic behavior Princeton University Press, Princeton, N.J, in: Ormala 1986
Nonaka, I., Takeuchi, H., 1995, The knowledge creating organisation, Oxford univ.press, New York NY
Ormala, E., 1986, Analysing and supporting R&D project evaluation, Espoo, ISBN 951.38.2611.2
Pahl, G., W.Beitz, 1996, Engineering Design - a systematic approach, Springer Verlag London Limited, ISBN 3.540.19917.9
Philipse (Herman), 2004, Belangeloos beminnen, Academische Boekengids 46, september 2004, UvA/RuG/UL/UU, pp 2-4
Rapoport, Anatoli, 1986, General System Theory, Abacus Press, ISBN 0856261726
Roozenburg, N.F.M. en J.Eekels, 1998, Productontwerpen, structuur en methoden, Lemma, Utrecht
Schön, D.A., 1987, Educating the reflective practitioner, Basic Books, New York
Shannon, C.E., W.Weaver, 1959, The mathematical theory of communication, University of Illinois Press, Urbana Ill., in: Fauconnier, 1986
Simon, H., 1996, The sciences of the artificial, MIT Press
Suh, Nam P., 1990, The principles of design, Oxford University Press, ISBN 0.19.504345.6
Ulrich, K.T., S.D.Eppinger, 2000, Product design and development, McGraw-Hill international editions, ISBN 0.07.116993.8
Wasserman,S., K.Faust, 1994, Social network analysis: methods and applications, Cambridge University Press, Cambridge, ISBN 0.521.38707.8
Wheelwright, S., K.Clark, 1992, Revolutionizing product development, Free Press, New York
Yoshikawa, H., 1987, General design theory as a formal theory of design, Intelligent CAD, vol1, North Holland, Amsterdam
Zwaan, A.H. van der, J.M.L. van Engelen, 1994, Bedrijfskundige Methodologie [in Dutch]: een wetenschapstheoretische context, p. 9, Bedrijfskunde, vol 66, nr. 1