The Pattern Language Network

Taming web2.0 in Higher Education

Validity, Resonance and Aggregation

Posted by yishaym on August 21, 2008

I’ve been chatting with Christian Kohls of Knowledge Media Research Center about pattern languages, workshops, community engagement, and the big picture. The discussion brought up some issues which where floating around EuroPLoP, and resonated with recent discussions on the Liberating Voices mailing list (pss – the book is out).

I see three major challenges pattern language communities to address:

Validity

What’s the scientific process of showing that a pattern does what it claims? Is it science, or is it art? What kind of evidence does a pattern need? How can we get the scientific community to accept patterns as a valid tool of knowledge production?

Even looking at Christopher Alexander’s patterns, the question arises. Alexander has a “confidence” measure, but what is it based on?

Resonance

In the Learning Patterns project, we noted:

Paradoxically, often as more expert knowledge is embedded in a pattern language it becomes less accessible to novices. The Learning Patterns project has tried to address this issue by a small set of Trails which accompany our pattern language.

But perhaps the problem goes deeper. Again and again, at every workshop we run, we see how hard it is for people to “get into the pattern groove”. Primarily, patterns are about abstraction without loosing context – and I think that is precisely what most people find hard. No wonder patterns have caught on so well in software engineering communities. After all, abstraction in context is what software engineers are trained to do.

So how do we break out of the cosy cult of patternisers, and make the knowledge we accumulated accessible to the wider public?

Aggregation

Once there was one pattern language for architecture (Alexander’s). Then there was one for software (GoF). Now there’s hundreds. Spread all over the place. Patterns are supposed to capture the essence of a recuring problem and its tried and tested method of solution. But they are supposed to capture it ONCE. In a way that can be referenced, linked, composed into larger structures or decomposed into sub-elements. What we’re seeing now is a fragmentation which defies the core purpose of the project.

How do we avoid reinventing wheels? How do we make sure that we build on each others’ work as much as possible, and aggregate design knowledge systematically?

Advertisements

8 Responses to “Validity, Resonance and Aggregation”

  1. Christian Kohls said

    Hi,

    to extend Yishay’s thoughts and maybe launch a discussion, I would like to add some comments on validity, usability, and scientific value/progress.

    To begin with, we want to find patterns that are valid in a sense that they exist in reality. We want to have patterns that we have *seen* several times and not only imagined. That makes a pattern valid in a sense that its structure really exist. The reoccurrence of structures can be tested in the same way as other scientific laws. But there seem to be at least two more requirements to pattern validity: wholeness and quality. Wholeness means that you decompose the structure in a way that each part is a gestalt that can exist on its own. For example, you can decompose a bike into brakes, wheels etc. But just taking, lets say, a quarter of a wheel does not make sense. Yes, the quarters of wheels are recurrent, but having the quarter of a wheel is not a form of wholeness. Therefore, I would call the quarter of a wheel an invalid design pattern. Some people say that wholeness and “quality without a name” are the same (e.g. Patterns Oriented Software Architecture, volume 5). But I don’t think so because there are forms that are whole but that do not bring quality to me. Weapons are an extreme example. I can see the gestalt of a weapon but I cannot see its quality. But people have different views on that issue. By definition, design patterns generate values because they solve problems. However, values are dependent on personal beliefs and opinions (although Alexander thinks different of that). What does that mean? Well, recurrence and wholeness are matters of testable validity, but the generated values are more difficult to judge on. The best you can expect is probabilistic validity – meaning that a given form A is more likely to be a fitting solution than form B for most of the people. But you will never get all of them. To resolve this conflict, let’s remember that there is context as well. Context could address the value system the pattern is based on: “If you are that kind of a person, than this pattern could work for you…”. Unfortunately, this usually does not happen. In fact, there are a couple of Alexander’s patterns that I do not like. I appreciate that the solutions he proposes will work for a lot of people but I am not one of them. In the universe of Alexander I do not exist – because he claims that his patterns work for ALL people and that beauty is empirical testable as any law of nature. This claim, however, makes the patterns in question invalid from a scientific point of view – because I do exist. And if there is only one existing person who does not feel the quality, then the quality cannot be a law. Summary: Validity for a pattern comes from recurrence, wholeness and the right scope of applicability – including personal preferences.

    Besides validity there is usability. A pattern can be valid but useless. It can be useless because it is so trivial that everybody else perfectly knows the solution as well. It can be useless because it addresses a problem so specific and rare that documenting it is not worth the effort. It can be useless because it is written badly, i.e. bad story telling, bad structure, missing description fields or use of inappropriate language. It can be useless because better solutions have been found already. Also, its usability can suffer from the wrong levels of abstraction, granularity and details. Note that these three issues will not affect the validity of the pattern but only its usability. Describing the pattern VEHICLE rather than BIKE is not wrong. It is just that describing a BIKE is more likely to be useful. You can also describe bikes in all of its details, e.g. enumerating all possible sizes, colours, gears etc. – giving too much details does not harm validity but it does not really help our understanding. Decomposing the bike down to the screws is valid as well since screws are whole things – but again usability is harmed. Since design patterns are, by definition, supposed to be generative, one could argue that a design pattern that lacks usability and usefulness is invalid.

    Finally, let us consider scientific value. One could have a fine pattern, perfectly valid and addressing difficult problems, and still gain no scientific value. To understand this, consider Newton’s laws. These laws are valid (conditions apply…) but documenting them again would bring no scientific progress in physics. However, if you document it in a new way that makes the law easier to understand for students, it can mean scientific progress in the field of education. Recognizing that one method of explaining is better than the other extends our knowledge. It is the same with patterns. If you are the first to abstract from several recurrent design forms then you have new scientific findings – a theory about the regular structures of a thing or process. If you find better ways to document patterns already known, you add practical value. Also, to find that pattern description A is more helpful than pattern description B adds scientific value but of a different kind.

    Well, these are my current views on validity, usability and scientific value. These views are subject to discuss. These are views rather than definitions or statements. To make more helpful definitions I would love to refine my views and see if they are shared or rejected by other pattern peoples.

    I think we are still in the discussion about many things, and we are far from saying what is right or wrong. The only thing that would be wrong is to claim that we share already the same understanding of patterns. No, there are many flavours to it. I remember a vivid discussion I had with Till at EuroPLoP08 whether a TABLE is a pattern or not. For me, a table certainly is a design pattern: it is recurrent and solves problems in a perfect way. Of course, to document and describe a table has no values at all: neither practical (what would you learn from it?) nor scientifically (what’s new about a table?). But: Just that everybody knows that an apple drops down does not mean that the law behind it is no longer a law, or does it? But, as I have written above, lack of usefulness could make a pattern invalid – depending of one’s view. Therefore, saying that tables are not patterns is not wrong at all. But then, in a world with many badly designed tables how can it be useless to remind the engineers to creates tables of the right heights and locating the legs in a way that does hurt…
    So, I guess that table discussion will continue in the future and I enjoy it very much. 🙂

  2. I’m glad to see this work! I looked at the idea of pattern language validation in my Liberating Voices book that Yishay mentioned. Although I’m not so sure that a pattern language can be “proved” to be correct,I came up with something that I think is (hopefully) a good start. This image below shows the relevant “components” that should be considered. I’d maintain that it’s through the correspondence between the components that the validity or legitimacy would be established.

    (The image is at http://www.publicsphereproject.org/patterns/pattern-language-validation-components.png)

    I’d like to think that the image was self-explanatory but I have a feeling it’s not…

    PS. I’m also interested in the integration of the various efforts. Our system has an XML interface — not working however since we moved to a new ISP. It should work soon. I’m hoping we can tweak that somehow.

  3. Stewart Dutfield said

    Yishay,

    Regarding validity, Alexander considers himself a scientist, and his own work rigorously empirical. The objective claims he makes for his constructs are grounded in his observations of things natural and built, and some of which he considers counter-intuitive. In a recent summary of Empirical Findings from The Nature of Order, Alexander sets out a number of steps in the books’ thesis that he considers objectively demonstrated.

    Perhaps it is easier to show the difference between life and not-life in buildings and artefacts, as Alexander claims to have done, than in more complex social entities. Establishing the link between patterns and this quality of life is, as you imply, also tricky.

  4. Christian Kohls said

    @Doug: The web marketing works :-). I am now curious to read the book and learn more about the graphic you linked. I think that clarification about the parties, entities, artefacts (both the objects and processes that manifest the patterns and the pattern descriptions), and their relations is overdue.

    @Steward and Yishay: Is a pattern science, or art?
    Personally I consider patterns as a “science of the artificial” (linking it to the thoughts of H.A. Simon). Mining the patterns certainly can be done in a scientific way. For me, mining a pattern is to build a theory about the invariant properties and relations of a class of design artefacts (includinging both artificial objects and artificially created processes). Such a theory is about the nature of form and quality of objects in that class. Like any theory it can be valid or not.

    Applying the patterns (conciously or not), however, is an art. Describing the patterns in a way that makes them useable is art.

    Hence, writing patterns is an art, the “pattern language of pattern writing” is a theory (a good one indeed, as I have learned so much from it).

    Research on the application of patterns would be science again. Also, research on how patterns emerge (e.g. piecemeal growth, or deductive reasoning) is science.

    Maybe this strong relationship of science and art makes patterns so sexy for us? But it also gives as a hard time, because the scientists tell us that our patterns are too practical and the artists tell us that our patterns are too theoretical…

  5. yishaym said

    Wow. what a discussion..

    Stewart: thanks for the reference. I’ve added it to our bibliography. Three notes:

    Alexander himself declares: It is also clear that more rigorous experiments along the same lines can be done, with larger samples, to reach conventional standards of scientific acceptance. I have not pursued this traditional scientific avenue to its full conclusion, since the construction of the logic of this chain of reasoning was a harder and more important task, arduous in the extreme. […] I look to my colleagues and to a new generation of scientists to carry this work forward with the necessary rigor.

    Alexander’s method is phenomenological. By all means a valid Philosophical approach, but one that may be contested as an empirical method in the sense of natural sciences.

    I haven’t read order of nature. I am only familiar with Alexander’s mathematics in Synthesis of form. I found it very poetic and beautiful, not necessarily rigorous.

    Doug: great image. indeed self-explanatory. however, I think it assumes a hierarchical relationship between pattern developers and users. In [1] we argue for a participatory approach. Lets think how your model can be expanded in that direction.
    As for integration and interoperability, have a look at the Pattern Exchange initiative. Although I should say – we’re happy to host your pattern on our system, if that helps.

    Christian: Interesting that you stress Wholeness. In Alexander’s recent work, that seems to come up as a unifying principle. First, in the paper Stewart mentions, and in the Living neighbourhoods project.
    Pattern are definitely an art, but I do believe they can, and should, also acquire a scientific dimension.

    I think there’s a deeper issue underlying this discussion. I see design pattern as part of the paradigm of design science [2], in the tradition of Herbert Simon. With that perspective, what counts as scientific method?

    [1] Yishay Mor and Niall Winters (2008), Participatory design in open education: a workshop model for developing a pattern language. Journal of Interactive Media

    [2] Yishay Mor and Niall Winters (2007) Design approaches in technology enhanced learning.
    Interactive Learning Environments15(1):61-75

  6. […] but do we have any evidence, in the scientific sense? In a way, this brings back the discussion on validity, resonance and aggregation. Yes, patterns work great for those who believe in them. But shouldn’t we aim […]

  7. […] Comments An inconvinient thes… on Validity, Resonance and A…Eureka!, again … on “Eureka!” Sketchin…yishaym on How to write a […]

  8. Hmmm… this response is very late. Since I’m here I might as well send my thoughts along.

    Although I can see how the diagram I proposed could suggest a hierarchy, I don’t intend it that way. There are two reasons. First, the diagram includes a two-way relationship between pattern developers and direct pattern users. Second, and this is made clear in the prose in the book (and unfortunately not as clear in the diagram), the developers and direct users aren’t necessarily distinct. I suppose I could have shown those two circles as overlapping but the two *roles* are distinct, whether the people assuming the roles are.

    BTW, after eight years of pushing forward on this project, the book is definitely available. We see the book as a step along the way, not as the main objective of the project.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

 
%d bloggers like this: