The web-site is now in readonly mode. Login and registration are disabled. (28 June 2019)

Evaluating partial designs

Story: engineering setting, collection of requirements, collection of component known...

Cloud created by:

Jon G. Hall
26 January 2010

Story:
As a practicing engineer, Derek designs artifacts that satisfy an identified need. The need is not stationary, nor its context: both exist in the real-world. collection of requirements, collection of component known (architecture, pieces, structures) goal, new deaign, uncertian outcome, creativity needed, external person to design process who shares risk of failure, judgement how to compile components to satisfy requirements. Ask person: yes/no
satisfactory validation of design by external. Incremental building, nothing to show in intermediate to check with external if progress is satisfactory. Use expertise to break down problem, preliminary safety analysis, ask: is there any obvious reason why this partial design won't be good enough?

 

Context Engineering design project

Problem: How do you know if and in which way a problem/solution is good enough to proceed?

Forces:
external evaluator to project, project cannot be presented

Solution:
S1: preliminary safety analysis
S2: understanding why other related project failed, trace back reasons for failure

 

(Academic research and other stuff at link below:)

Extra content

Embedded Content

Contribute

nicole schadewitz
9:30am 28 January 2010


I think your observations might be quite relevant to design education. Often design students need to present their projects to externals. Often this only happens at the final design crit. But I was reading that the externals thought often it would have been better to involve them earlier in the evaluation of the student's design so they could have helped them and showed them a better way. At the end of a project this might be too late. I think building on an external evaluator’s knowledge is important in early stages of the design process. In my experience, there is no standard process in place, but the evaluator uses her or his expertise to steer the project in a better direction.

Jon G. Hall
9:39am 28 January 2010


You make me very happy with your comment:-) I'm interested  to work this up with you. Here's another paper, a technical report, that goes into more detail than the DRS paper linked to earlier: http://computing-reports.open.ac.uk/2008/TR2008_02.pdf

nicole schadewitz
9:52am 28 January 2010


Ok lets start with exploring the problem.

So the context in which this occurs is design engineering, and I have observed it in design education (architecture - it is a paper I need to find and will post it here in refs.)
Our question or problem statement is: Is there any obvious reason why a partial design won't be good enough (to develop into a full design)? This is your original question, right?

nicole schadewitz
1:13pm 28 January 2010


The refernece on design juries by Anthony includes findings that interim reviews help students more than the final crit.

Derek Jones
6:48pm 8 May 2010


Jon - very interesting papers which overlap with a number of things that I am thinking about at present, thanks.

There is one problem with the fractal view of design - it requires absolute linearity. Complex sets are an instance with a given set of starting parameters and they generate their geometry from this entirely. There are certainly analogies with design in this (e.g. design cycles within design cycles), but what this then misses is the fact that a single design cycle (or self-similar pattern) can affect the shape of other patterns. Fixed sets (like Julia or Mandelbrot sets) are complex through self-similarity in a linear way.

So, to take the analogy further, we assume that design follows a complex path but we cannot predict this path without knowing very precise starting conditions (one of the defining concepts of complexity).

Basically, we need several complex sets on top of one another, each with a different set of starting conditions that can interact with one another to 'collapse' into simplified true set.  

But - what if we look at it from the other side :

If we assume that we cannot know the starting condition to determine a complex path, we have enough maths now to be able to re-create certain other features of a complex set. Moreover, we could imagine certain 'fixed points' or other attractors in the map to be able to 'play' with it. If we try and visualise this using iteration and feedback then it might be possible to generate 'shapes' within the complex geometries that could be used to understand the nature and shape of the complex 'meta-map'.

To use another geometric analogy, if we have the original Lorenz system (and do not yet know the nature of the attractor in that system), we can still run iterative simulation in order to visualise the attractor. In fact, what we really want is to be able to start in the middle of the system, create data that shows how it behaves and try to work back to see if there exists any kind of attractor that can be followed.

Imagine what could be achieved if We Feel Fine was truly interactive - if we could manipulate the complexity of the data in such a visual way to be able to establish previously hidden patterns (this is something I am trying to work on just now, not much fun )...

Derek Jones
7:23pm 8 May 2010


Sorry - last post was an explosion of maths. This time, I will stick to design :)

I think Nicole hit it spot on with the architectural crit. reference (I'm an architect btw). As we focus down on a particular series of choices (read design) we automatically limit our thinking (by habit, codes of practice, constraints, experience, etc etc). It is usually only when an external source of information / inspiration comes along that we really re-evaluate these choices. Self evaluation is rarely carried out well since we are actually part of the process itself (imho).

So basically, we seek an evaluation process that is efficient and timed well to have a beneficial effect on the final design (or that part of the design). And, as Nicole re-states, we seek to ask 'is there a reason why this 'solution' cannot be developed further?'  

I would suggest that (to use Schon's terminology) the key to this is Framing (?), i.e. asking 'what is the actual goal?' Which in turns allows you to identify values against which you can measure the partial design. Again, this is often a stage that is actually missed in my experience - often replaced by continual playing instead of constructive designing.

Perhaps the key is not the answer (or partial answer), but the finding the 'good' question.

Jon G. Hall
9:21am 14 May 2010


Dear Derek,

Thank you for your email, and your kind words.

Your post is too deep for me to comment quickly on its mathematical detail, but I would like to start up this discussion with you; as you're at the OU, it might be over coffee in our new hub! (I'm at J.G.Hall@open.ac.uk.)

Assumption: I interpret absolute linearity to mean the causal dependence of any part of a process - other than the initial action - on some other part. It's like a sequential program.

Given this assumption, then I'd like to argue that there is no such constraint: our fractal view is only approximate (a better term might be recursive similarity, but fractal is more readily understaood), with patterns generating inner and outer structures. For those structures, we have defined a process algebra of possible design processes - that sounds so unattractive, doesn't it, but it gives us our approximate fractal - in which a design process is either a basic (or trusted, i.e.,  not in need of validation) process, or a sequential (operator ';', linear) composition of two other processes, or a parallel (operator '||', non-linear) composition of two other processes, or an (approximate) fractal (operator: '|><|' or bowtie) composition (see attached picture). Of these, the parallel composition might be that which removes the absolute linearity... as through it resources can be consumed independently of each other.

I have pictures, but don't know how to post them.)

I'd really like to discuss more with you, to test my assumptions, but also to learn from your perspective on design. There's much for us still to do! If we do manage to meet, I will summarise for this page.

Best wishes, Jon



Derek Jones
9:25pm 16 May 2010


Hi Jon - It would be great to have a discussion about this. Unfortunately I'm an AL so I don't get down to 'head office' too often :), but I would still be interested in hearing more. I will drop you an email in the next few days.

I do see what you mean about recursive similarity and I also see what you mean about 'like fractals'. Would another analogy be the basic principle of image compression algorithms? (i.e. self similarity in places where this is appropriate and other self similarities (or even bespoke processes) in others ?)

I would completely agree that a parallel operator would offer the most obvious 'non-similar' route, but a composition could also exhibit this (particularly if you allow compositions of compositions (g(f(x)) type of thing) since alternative combinations of compositions could even 'evolve'. 

I guess the 'big' question, then, is 'is this computable' (in the sense of P vs NP... :D)? 

Jon G. Hall
6:23am 21 May 2010


Derek,

I'm jon_hall on Skype, should that work for you

I look forward to hearing from you.

Best wishes, Jon

Contribute to the discussion

Please log in to post a comment. Register here if you haven't signed up yet.