Print

Print


The other point is that it is fascinating that this stream is continuing to generate such interests...

Roger
On 8 Feb 2008, at 20:55, Roger Harnden wrote:

Rod, to the point as always.

My point was that the VSM might not be the tool to analyse the problem - it was not that I am happy with accidents that might include myself among others.

There are many strands going on in this thread. There are observer determined points, there are SSM points, there are 'hard' and 'soft' points.  As has been said before, the VSM is not a universal panacea. The VSM is not a 'tool for stopping air crashes'. It never was, and it never will be. Human beings and 'messy' systems are involved, as well as technical ones. The VSM might well help highlight the overall situation. It will not in itself give an answer. It is not a design blueprint for the 'perfect automated system'. That's all I meant to say.

Roger


On 7 Feb 2008, at 22:58, ROD THOMAS wrote:

Hi Nick,
 
Yes, just dipping in and out of this thread over the weeks has revealed quite a great deal of confusion as to just what is offered by cybernetics to the prevention of failure. As if it offers the magic spell to ward off all harmful possibilities!
 
As I understand it, 'stability' is the cybernetic term for the output (that is of interest to us) remaining within acceptable bounds. Notwithstanding Roger's implied observation that from an ecological viewpoint an air traffic crash may be no bad thing, I think most people's intentionality to the world would call it a 'disaster' or 'accident'. Hence, to those people, cybernetics would say that 'stability' can sometimes be achieved by what Stafford sometimes called 'implicit control' - fast action, continuous and ideally automatic, negative feedback. As I undersatnd it, this is what the Wright brothers achieved - they designed an aircraft that could not fly by itself, instead they introduced a pilot to offer feedback adjustments that counteracted tilt or dip. As De bono once wrote, 'their eyesight and seat of pants' completed the feedback loop. However, that cybernetic advance did not overcome the many disturbances to flight that are not overcome by eyesight and seat of pants.
 
Ultrastability still relies on feedback, but as I understand it, its where there are a number of interacting feedback loops that continually act to reconfigure until all sub-sytems are within stable zones: an equilibrium for the system as a whole. This means that control may not be located in a single controller - simply monitoring horizon and seat of pants - it may be distributed throughout the structure of feedback relationships. Hence a disturbance to any one system, potentially regardless of cause, will result in a series of changes that have no end until the whole system recovers an equilibrium state. This was Ashby's machine - strangely enough built from surplus RAF equipment. No doubt modern aircraft have these kinds of arrangement: with all their red warning lights etc.
 
But obviously (?) even ultrastability can't thwart a devastating missile or a bomb that destroys the homeostatic configuration.
 
I remember Stafford used to talk about Ashby's Law and airport security - his example was that his cigar case was an imaginery bomb and no-one once looked in it when he checked in at the airport. What are we to do - go through security bollock naked? But as every special forces soldier knows - the body itself has one or two natural suitcases.
 
So in short, we are (wo)men not gods.
 
Rod Thomas

Nick Green <[log in to unmask]> wrote:
Dear Paul

How about this:

Ultrastability is a desideratum of a system seen as a number of interacting
homoeostats. Clearly any perturbation if big enouigh will destroy coherence.
A large enough meteor crashing on earth for example could end
mammalian/human life.

Ashby set up a technical (notationally over rich perhaps) description of
ultrastability in his "Design for an Intelligence Amplifier" (Automata
Studies ed Shannon and McCarthy Princeton UP 1956) and embodied in the many
stable states achievable in his hardware homeostat (the step function of
which we may see as System 4) in his "Design for a Brain". There is a
feeling that redundancy is important, reflected in Ashby's Law of experience
(and his technical idea of cylindrance isomorphic, perhaps, to Stafford's
paradigm of the logical search in Brain).

But having said all that I'm not sure we can say much with certainty about
the future. A number of small unexpected perturbations all within bounds
might defeat any control policy. With adequate Variety engineering at least
we can monitor what we know to be these critical variables and their bounds,
simulate worst cases and deal with problems as they arise in as timely a
manner as possible -but that's probably more than most would say. We laugh
at the obvious errors made by the "Wobbly Bridge" designers but can we
honestly say we can produce designs that will never go into destructive
oscillations (like the sub-prime credit errors threatens to)?

However there are simple fundamental checks at present that are not done and
these we can tackle with some certainty. What is the flux of CO2 over
desert, sea, pampas and rain forest? We dont know. What was the cost of risk
in sub-prime lending? We didn't know. What are the daily outcomes of
medicating patients again we don't know- but we could know. All we can do
(Chaitin like) is decrease the probability of halting (or going extinct,
non-viable) by adding variety (men, machines, money) where our quantitative
models, always improving, suggest it is most needed. In effect from VSM we
set up a transparent structured heuristic for survival.

Incidentally if anybody wants a textbook on Risk I have been using Bedford
and Cooke "Probabilistic Risk Analysis: foundations and methods" (Cambridge
UP 2001) for some years now and, at least, it makes me feel better.

I once asked a Chem Eng friend who had been doing the Risk analysis for
Sizewell B what he did. "Oh", he said "you know". Well I didn't and that is
why I asked but it turns out they just looked at scenarios and probabilities
(sometimes of rare events - which can be tricky); added some (for logical
OR), multiplied others (for logical AND) and answered questions like "the
chances of aircraft dropping out of the sky onto the reactor", the pressure
vessel failing and the way a smoke plume would drift. That is one way of
simulating future worst cases and hence managing the future. System 4 is
asked is the containment vessel strong enough? What do we do if it isn't?
What are the chances of it failing due to excessive perturbation in, say, 25
years?

Best

N.

----- Original Message -----
From: "Paul Stokes" <[log in to unmask]>
To: <[log in to unmask]>
Sent: Friday, January 18, 2008 1:43 AM
Subject: Fw: System failure


> Arthur,
>
> It is my understanding that for a well-designed cybernetic system you do
> not
> need to specifiy in advance causes of possible future disturbance to the
> system.
>
> It would be a very interesting exercise though to specify an ultrastable
> (Ashby) aircraft capable of dealing with any possible source of
> disturbance.
> Sounds impossible? Any takers?
>
> Paul
>
>>
>> ----- Original Message -----
>> From: "Arthur Dijkstra" <[log in to unmask]>
>> To: <[log in to unmask]>
>> Sent: Thursday, January 17, 2008 4:52 PM
>> Subject: Re: System failure
>>
>>
>>> Thanks Stuart and all,
>>> Yes I have read the Perrow's book. Because of complexity and coupling we
>>> can
>>> expect failures. I the safety management system (SMS) these failures
>>> should
>>> be anticipated and avoided or controlled. I want to work backwards, so
>>> from
>>> the accident, via the conditions into the organisation to find
>>> precursors
>>> and control them. The way you understand accidents shape the way you try
>>> to
>>> prevent them. For now I want to describe accident in cybernetic
>>> language.
>>> Regards,
>>> Arthur
>>>
>>>
>>> -----Oorspronkelijk bericht-----
>>> Van: Forum dedicated to the work of Stafford Beer
>>> [mailto:[log in to unmask]] Namens Stuart Umpleby
>>> Verzonden: donderdag 17 januari 2008 17:25
>>> Aan: [log in to unmask]
>>> Onderwerp: Re: System failure
>>>
>>> Probably you know about Charles Perrow's book Normal Accidents, 1984.
>>> As I recall, he claims that if the number of elements that can fail is
>>> large and the interconnections among elements is large, occasional
>>> failure is "normal." Stated differently, complexity can be a cause of
>>> failure. Back up systems prevent a crash due to the failure of a
>>> single component. Hence, several things need to go wrong at the same
>>> time to cause a crash. So, one looks for combinations of failures and
>>> factors which cause several components to fail at once. Perrow's book
>>> was widely read in the months and years before y2k.
>>>
>>> On Jan 17, 2008 9:39 AM, Arthur Dijkstra <[log in to unmask]>
>>> wrote:
>>>> Hi Frank and others,
>>>> Thanks, I am aware of this. The challenge is to relate data from the
>>>> operational flights and the organisation to the probability of a
>>>> accident.
>>>> Therefore I need a exhaustive list of possible ways to crash a aircraft
>>> from
>>>> a cybernetic perspective.
>>>> Regards,
>>>> Arthur
>>>>
>>>> -----Oorspronkelijk bericht-----
>>>> Van: Forum dedicated to the work of Stafford Beer
>>>> [mailto:[log in to unmask]] Namens Frank
>>>> Verzonden: donderdag 17 januari 2008 15:33
>>>> Aan: [log in to unmask]
>>>> Onderwerp: Re: System failure
>>>>
>>>>
>>>> Dear Arthur,
>>>> whilst this is not a cybernetics approach I think it could be useful.
>>>> It's
>>>> more front line but tells its own story..
>>>>
>>>> Extract from article
>>>> [...] But with so few crashes in recent years, air carriers and
>>>> regulators
>>>> have been trying to find other ways to identify potentially dangerous
>>>> trends. Instead of digging through debris, they now spend far more time
>>>> combing through computer records, including data downloaded from
>>>> thousands
>>>> of daily flights and scores of pilot incident reports.
>>>>
>>>> The information is stored on banks of computers, such as the server
>>>> housed
>>>> in a windowless office of a US Airways hangar here. Like its
>>>> counterparts
>>> at
>>>>
>>>> other carriers, a small team of pilots and analysts sift through
>>>> thousands
>>>> of records daily looking for the seeds of the next big air crash.
>>>>
>>>> In recent years, the team has uncovered such potential safety problems
>>>> as
>>>> unsafe landing and takeoff practices and difficult landing approaches.
>>>> The
>>>> data have helped pinpoint areas that pose an increased risk of midair
>>>> or
>>>> ground collisions and have led to the discovery of a large bulge in the
>>>> runway of a Vermont airport. Even after threats have been reduced, US
>>>> Airways' executives and pilots say they keep monitoring the data to
>>>> ensure
>>>> that their new procedures work.
>>>>
>>>>
>>> http://www.washingtonpost.com/wp-dyn/content/article/2008/01/12/AR2008011202
>>>> 407.html
>>>>
>>>> Hope this helps.
>>>>
>>>> Regards
>>>>
>>>> Frank Wood
>>>>
>>>> ----- Original Message -----
>>>> From: "Arthur Dijkstra" <[log in to unmask]>
>>>> To: <[log in to unmask]>
>>>> Sent: Thursday, January 17, 2008 2:10 PM
>>>> Subject: System failure
>>>>
>>>>
>>>> > Dear all,
>>>> > In my project to develop a Safety Management System for aviation I am
>>>> > evaluating different categories to describe aircraft accidents. Using
>>>> > cybernetics, I want to make a exhaustive and usable list of the way
>>>> > an
>>>> > aircraft can crash. Sort of 50 ways to crash your aircraft :-) Usable
>>>> > means
>>>> > in this context that in an organisation events can be related to the
>>>> > possible accidents. As a cybernetician how would you build such a
>>> category
>>>> > (hierarchy of categories) to describe the possible accident types ?
>>>> >
>>>> > Thanks for your response,
>>>> > Arthur
>>>> >
>>>> > For more information go to: www.metaphorum.org
>>>> > For the Metaphorum Collaborative Working Environment (MCWE) go to:
>>>> > www.platformforchange.org
>>>> >
>>>> >
>>>> > ---
>>>> > avast! Antivirus: Inbound message clean.
>>>> > Virus Database (VPS): 080116-1, 01/16/2008
>>>> > Tested on: 1/17/2008 2:16:49 PM
>>>> > avast! - copyright (c) 1988-2008 ALWIL Software.
>>>> > http://www.avast.com
>>>> >
>>>> >
>>>> >
>>>>
>>>> For more information go to: www.metaphorum.org
>>>> For the Metaphorum Collaborative Working Environment (MCWE) go to:
>>>> www.platformforchange.org
>>>>
>>>> For more information go to: www.metaphorum.org
>>>> For the Metaphorum Collaborative Working Environment (MCWE) go to:
>>> www.platformforchange.org
>>>>
>>>
>>>
>>>
>>> --
>>> Stuart Umpleby, Research Program in Social and Organizational Learning
>>> 2033 K Street NW, Suite 230, The George Washington University,
>>> Washington, DC 20052
>>> www.gwu.edu/~umpleby, tel. 202-994-1642, fax 202-994-5284
>>>
>>> For more information go to: www.metaphorum.org
>>> For the Metaphorum Collaborative Working Environment (MCWE) go to:
>>> www.platformforchange.org
>>>
>>> For more information go to: www.metaphorum.org
>>> For the Metaphorum Collaborative Working Environment (MCWE) go to:
>>> www.platformforchange.org
>>>
>>
>
> For more information go to: www.metaphorum.org
> For the Metaphorum Collaborative Working Environment (MCWE) go to:
> www.platformforchange.org
>
>
>
> --
> No virus found in this incoming message.
> Checked by AVG Free Edition. Version: 7.5.516 / Virus Database:
> 269.19.5/1228 - Release Date: 16/01/2008 09:01
>

For more information go to: www.metaphorum.org
For the Metaphorum Collaborative Working Environment (MCWE) go to: www.platformforchange.org
<[log in to unmask]><[log in to unmask]><[log in to unmask]><[log in to unmask]><[log in to unmask]><[log in to unmask]><[log in to unmask]>

For more information go to: www.metaphorum.org For the Metaphorum Collaborative Working Environment (MCWE) go to: www.platformforchange.org




For more information go to: www.metaphorum.org For the Metaphorum Collaborative Working Environment (MCWE) go to: www.platformforchange.org




For more information go to: www.metaphorum.org For the Metaphorum Collaborative Working Environment (MCWE) go to: www.platformforchange.org