Human Behavior in System Dynamics Models

This forum contains all archives from the SD Mailing list (go to http://www.systemdynamics.org/forum/ for more information). This is here as a read-only resource, please post any SD related questions to the SD Discussion forum.
Locked
"George Backus"
Member
Posts: 23
Joined: Fri Mar 29, 2002 3:39 am

Human Behavior in System Dynamics Models

Post by "George Backus" »

ads.com>
Subject: REPLY Human Behavior in System Dynamics Models (SD3337)
Date: Tue, 24 Jul 2001 15:01:40 -0600
Organization: gabackus@home
MIME-Version: 1.0
Content-Type: text/plain;
charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
X-Priority: 3
X-MSMail-Priority: Normal
X-Mailer: Microsoft Outlook Express 5.50.4522.1200
X-MimeOLE: Produced By Microsoft MimeOLE V5.50.4522.1200

Jack Ring asks about two very different situations and the applicability of
SD in addressing them. The first is a straightforward problem of dealing
with firings based on "personality conflicts." The second is a more
problematical exploration of worker enthusiasm. Before I head into this, I
feel the irrepressible need to introduced it with (another) background
harangue on the SD approach.

SD is a problem-oriented approach. Its goal is not to model the system but
to only include those aspects of the system important to understanding and
resolving the problem. This relates directly to the insight of another
recent listserve thread concerning the order of systems. The equations are
not there to describe assumed results but rather to explain the causes of
the results - and thereby provide the opportunity to intervene and produce
more desirable results.

While the understanding of the physical and financial aspects of a system is
much more uncertain and limited than we would often like to acknowledge, the
understanding of human behavior is even more ambiguous. If the human
behavior either does not produce dynamics relevant to the problem solution
or can be treated as a relatively fixed, uncontrollable entity, then it can
be subsumed in the mathematical structure of the model. Many problems
actually conform to this situation and human elements (the humans having the
problem) over-rate their importance to the underlying dynamics. If the
human behavior is surmised as the cause of the problem or an integral part
of it, then those relevant aspects of human behavior do need to be
simulated.

All science is inexact. Each day, new more refined experiments discover new
unexpected features (mechanisms) in cosmology, quantum mechanics, genetics,
etc. Occasionally, the new findings lead to an overthrow of our
understanding of some aspect of reality (See Stermans human-behavioral
modeling of Thomas Kuhns Structure of Scientific Revolutions.)
Nonetheless, the old more aggregate understanding was often useful to
address pressing problems (for example, Newtonian physics versus Relativity
and classical physics versus quantum mechanics). The issue is not fully a
question of right or wrong but one of rationality. Is there a causal
(hypothesized) explanation of responses (cause and effect) consistent with
measured conditions? If the result cannot be quantified and measured - even
qualitatively, then it is irrelevant (useless) for our purposes and
irrational from our paradigms perspective.

Physical and financial dynamics have a limited number of viable hypotheses.
Often all of any conflicting explanations can be "included" in the analysis
process. The need for self-consistency (uniform rationality) and the
comparison of simulated output to simulated output often falsifies (see Karl
Poppers The Logic of Scientific Discovery) a large number of the competing
explanations. The remaining explanations can be represented as alternative
scenarios that cannot be discarded. We find that multiple self-consistent
presentation often lead to the same policy conclusions. Econometricians
call this encompassing. By this they mean there is a more generalized
data-generating process (DGP) that could be recognized/discovered or that
the various "perspectives are basically describing the same process but
from a different perspective. When the alternative perspectives do not lead
to the same solution, the perspective most consistent with the stakeholders
paradigm seems to usually wins out. Some things in life must remain
uncertain.

Human behavior adds further dimensions to the complexity of modeling. There
are too many degrees of freedom given the data and much remains to be
understood about even generic aspects of human behavior. We are still in
the Dark Age of cognitive/emotive science. We are left to assume that
either what we do not fully understand is irrelevant or we can do the best
we can to rationally include it. This "best" provides the best chance of
successfully solving the problem. The "best" may be far removed from the
ever-elusive "truth." As was the case in early scientific discovery,
aggregate perspectives (thermodynamics, for example) provide a means to deal
with the aggregate phenomena of concern. We do not need to know the
dynamics of individual atoms. We only need to have a relevant understanding
of the aggregate behavior relevant to our problem. Fortunately, in social
science, if our knowledge is limited to aggregate problems (those that are
big enough to impact us) so too is our cognition of underlying detailed
dynamics. We have an excuse to only deal at the aggregate level because
that is all we can perceive. I often use the concept of coping-skills to
contain these aggregate mechanistic behaviors. Coping skills are explicit
in the SD work of Ulrich Goluke on alcoholism and implicit in the expansive
SD work of Jack Homer and others on worker burnout. This aggregation
produces results consistent with experienced human behavior (therefore
useful) and provides a tractable means to test policy (therefore efficient).

With this background, I can then address the points Jack brings up. The
first issue is the corporate and economic loss (waste) associated with
"firings" based on personality conflicts. (I have to believe that some
graduate student or consulting firm has simulated this HR issue already but
need some other SD practitioner needs to help me out on that point.) The
first value of SD is to recognize the system context of the "problem." From
a system perspective, either there is a management personnel problem or
there is a position problem. If I am in charge of a suicide-prevention
hot-line or an air-traffic controller shift, the personality traits are
important. If the problem of under-staffing becomes critical, I need to
either change the definition of the position to allow more flexibility in
staff psyches (as in the tiered help-line services for many software
companies) or train (enhance the coping skills of) the employees before they
are placed on the front line. Properly designed field-tests or surveys can
determine the (relative) validity of the dynamics assumptions related to the
problem and proposed intervention.

If the problem is with a manager or management practice, then one or both of
those two issues are readily addressed. Management practice is readily
surmised using panel data from other firms. Many consulting firms have
"best practice" divisions designed solely to recognize the organizational
structures that work for given conditions. It still, however, remains the
domain of the system dynamists (my biases are showing) to give a casual
explanation of why these practices produce the results. If the problem is
with the managers who do the firing, the "coping skill limitation" is on the
managerial end. The manager may feel threatened, overwhelmed, victimized,
or simply powerless. Interviews with both thriving and ineffective managers
can help develop causal hypotheses to test via simulation. Additional
training, re-specification of the position, separation of the position into
additional hierarchies, aggregation/elevation of departments to a more
viable entity, etc. all come to mind. Each of these can be tested using the
best understanding of cause and effect against the evidence. (SD quackery
can misrepresent this approach when "practitioners" include correlation
equations producing dynamics rather than restrict themselves to the use of
only causal structures.) I would argue that the causal use of SD is
equivalent to respected historical-study approaches where value is derived
in producing a consistent and useful story. SD, in practice, is a language.
Any causal explanation (story) can be readily represented in an SD model.
There is no limitation for modeling human behavior.

The second issue Jack brings up is workforce motivation. The problem here
is that "work force motivation" is not a problem statement. It may or may
not be a contributor of a problem (recent studies indicate dissatisfied
workers are more productive: Lavis, C.A. & Sinclair, R.C. Are happy workers
better workers? Examining the effects of mood on productivity. Proceedings
& Abstracts of the Annual Meeting of the Midwestern Psychological
Association, 71, 11, 1999). If we are in the realm of SD, the issue is, for
example, productivity.

SD does not find mechanisms but tests hypothesized mechanisms. Humans
cannot (yet) relinquish their responsibility for their ignorance. (Many)
humans tend to be creative and can readily posit alternative "causal
explanations." The SDer can then map and reconcile these into a
self-consistent framework as noted above. The resulting self-consistent
causal representation (model) is then the best hope for rational policy
development. To be concrete, the "mechanism" Jack presents would appear to
fall into the conceptual domain of coping skills. When coping skills are
not nurtured or challenged, they atrophy and coping skills decline. When
overpowering forces prevent compensating actions, coping skills decline.
Appropriate challenges and discretionary flexibility reinforce coping
behaviors. Personnel in repetitive processes often have no discretionary
capabilities -- they can only follow the inflexible rules. In many of our
projects, we find these "line" workers to be much more stressed than senior
management who are supposedly dealing with the crises that made them so
desperate that they reluctantly resorted to calling us. Individuals tend to
operate near the left-side of the peak of their coping capability. The
approach is then to increase the coping skill level. This may include
reward systems that make failure less powerful than a success (as in the
case of R&D efforts) or it may require changes to the workplace that
reinforces self-actualizing behaviors. I am sure there are many other
causal possibilities, but the key is that SD is fully capable of modeling
the problem and the relevant aspects of the system containing it.

Jacks concluding remarks indicate that the focus of his interest is in group
dynamics, project work climate, organizational cultures, and policy setting
as a function of team selection. Organizational learning is another growing
aspect of SD efforts and possibly other listserve participants may directly
contact Jack on their efforts. Similarly, work-climate is a significant
focus for many consultants on the listserve. Basic group dynamics and human
behavior are often put in the domain of action science and action research
(I am just newly aware of this myself). See, for example,
http://www.actiondesign.com
esources/ and
http://www.actionscience.com/actinq.htm This area is related to Senges work
on the Fifth Discipline.

Lastly, I just want to comment on the "team selection" process. I think it
needs again to be in the context of a problem and that a generic
representation of "team selection" will remain elusive. Selecting team
members based on the problem being addressed has relevance and a definable
scope. The denotation of simply "team selection" is a general "SYSTEM"
issue and the SD paradigm argues that neither it nor any other paradigm can
give meaning to the concept of "modeling a system."

Another unfortunate soapbox episode thus ends.. I need a life :-)

George

George Backus
Policy Assessment Corporation
14604 West 62nd Place
Arvada, CO 80004-3621
Bus: 303-467-3566
Email: George_Backus@ENERGY2020.com
Paul Martin
Junior Member
Posts: 11
Joined: Fri Mar 29, 2002 3:39 am

Human Behavior in System Dynamics Models

Post by Paul Martin »

A couple of years ago we embarked on a thinking experiment, forming part of a
number of projects looking at strategies for sustainability, where we tried to
create a systematic framework for understanding resource use decision making
at three levels - the individual, the organisation and society- with
interactions between the three.

We eventually found, drawing on a great wealth of literature, some core themes
and repeated patterns, with in particular the sense that there are only two
flows within the systems - flows of information and flows of resource
(including economic flows) and there is, moderating the operation of all of
this, a system of values plus information ordering (decision structures) which
is multi-dimensional and which shapes the way in which these two flows
operate.

If anyone is interested you can obtain a partial summary of this in the form
of a pdf file, headed "Sustainability, strategy and society" from the download
section of our website at
www.profitfoundation.com.au.

I am not putting this forward as a definitive model, but it is at least an
integrative approach which overtly draws on the concepts of behavioural
explanation used in economic, sociology, law and social psychology into a
rough systems framework.

Any better integrative models would be greatly appreciated, especially since
whilst we would like to turn this into a SD model for teaching and analysis
purposes, we just do not have the resources (time and technical skill) to do
so.

n:Martin;Paul
tel;cell:0416015161
x-mozilla-html:FALSE
url:www.profitfoundation.com.au
org:The Profit Foundation
adr:;;;;;;
version:2.1
email;internet:paul_m@profitfoundation.com.au
note:Strategic advising for growth and sustainability
"Jay W. Forrester"
Senior Member
Posts: 63
Joined: Fri Mar 29, 2002 3:39 am

Human Behavior in System Dynamics Models

Post by "Jay W. Forrester" »

Are there any management or social system models that do not at least
implicitly include human behavior?

"Jack Ring" <
jring@amug.org> in SD3313 wrote
>Although the Sterman model to which you refer is quite interesting it is
>largely an economics model and includes no obvious variables for human
>behavior.

--
---------------------------------------------------------
Jay W. Forrester
Professor of Management
Sloan School
Massachusetts Institute of Technology
Room E60-389
Cambridge, MA 02139
tel: 617-253-1571
fax: 617-258-9405
From: "Jay W. Forrester" <jforestr@MIT.EDU>
"Jack Ring"
Junior Member
Posts: 12
Joined: Fri Mar 29, 2002 3:39 am

Human Behavior in System Dynamics Models

Post by "Jack Ring" »

I am not aware of any (but that is a <1% coverage likelihood).
The "implicit" part is the problem because the causal factors at the
people relationship level are not subject to validation nor calibration.

Jack Ring
From: "Jack Ring" <jring@amug.org>
"George Backus"
Member
Posts: 23
Joined: Fri Mar 29, 2002 3:39 am

Human Behavior in System Dynamics Models

Post by "George Backus" »

Jack Ring wrote

"The "implicit" part is the problem because the causal factors at the
people relationship level are not subject to validation nor calibration."

The only important issue for policy is the outcome of decisions. What people
think or feel is not relevant to the physical world, but how they act on
those values, feelings and thoughts is. The decisions people make are
readily measurable -- either from actual decisions or from surveys, polls,
or interactive simulations. A whole body of statistics deals with the
validity, applicability, and scalability of such information. This
information can then be used to rigorously simulate the decisions as a
function of perceived conditions and personal tastes, values, and
preferences. Qualitative Choice Theory is one of the more popular
approaches in that it is easily calibrated and parameterized, is robust, and
has the feature that allows "analogous" decision data to be applied to new
situations.

The work of the recent Nobel Prize winner Daniel McFadden is most
noteworthy. To start point to review this literature might be:

Structural Analysis of Discrete Data with Econometric Applications, Charles
Manski and Daniel McFadden (eds), MIT Press, 1986.

A basic starting text on a mathematical treatment of decision behavior might
be:

Decisions with Multiple Objectives: Preferences and Value Trade-offs. by
Ralph Keeney and Howard Raiffa, John Wiley & Sons 1976.

Historically many SD models routinely incorporated these knowledge areas in
the simulation to explicitly deal with human behavior.

G

George Backus
Policy Assessment Corporation
14604 West 62nd Place
Arvada, CO 80004-3621
Bus: 303-467-3566
Email:
George_Backus@ENERGY2020.com
"Jack Ring"
Junior Member
Posts: 12
Joined: Fri Mar 29, 2002 3:39 am

Human Behavior in System Dynamics Models

Post by "Jack Ring" »

Thanks, George, summarizing the SD view of modeling human behavior. I
do not question any of it. It is an excellent comparison point against
which hopefully I can be more clear about my point.

Perhaps you could tell me how SD could be used to examine and refine the
policies surrounding the well known statistic that 30% of "firings" are
due to personality conflicts, presuming, of course, that the objective
is to reduce that level of waste.

Or an easier one. Show me an SD model of a business process that
includes the "mechanism" that increases workforce enthusiasm. Please
highlight how the model and simulation serves to establish the policies
for that mechanism depending on whether the workforce is engaged in
repetitive production or expeditionary product development.

Please note that I am not taking issue with your summary. I am
attempting to clarify the aspects of human behavior and group dynamics
and project work climate and organizational culture that I have not been
able to find in the SD tribe. I am not saying that SD is in error. I
am seeking ways to solve real world policy-setting (team selection)
problems and am simply questioning whether SD has or can be made to
elucidate this problem.

Jack Ring
From: "Jack Ring" <jring@amug.org>
Bill Braun
Senior Member
Posts: 73
Joined: Fri Mar 29, 2002 3:39 am

Human Behavior in System Dynamics Models

Post by Bill Braun »

Jacks request for example models prompts me to wonder if weve lost the
notion of boundaries. If we are attempting to model the deep roots of human
feeling and emotion in the belief that they are the ultimate drivers of
behavior, it would seem to be a good time to propose we construct a model
of everything.

Jacks request also seems to strike at the heart of the implicit
relationship between SD, policy and problem focus. One one hand, given a
context (the problem) and a set of policies, we should be able to model
current reality. We can model the probability that people will adhere to
policies (a reflection of human behavior and itself a reflection of drivers
of human behavior) short of modeling the emotional system that produced the
behavior without taking anything away from the utility of the model.

On the other hand, it seems as though we would have to develop an SD model
of the brain to get close to Jacks request. I am as fascinated by the
prospect as I am puzzled by its relevance to SD. Just as Dr. Forrester has
admonished us not to try to model every facet of a system (the model of
everything), is it reasonable to accept that SD is not the tool for
everything without seeing that as a failure of the tool?

Bill Braun
From: Bill Braun <medprac@hlthsys.com>
Nelson Repenning
Junior Member
Posts: 10
Joined: Fri Mar 29, 2002 3:39 am

Human Behavior in System Dynamics Models

Post by Nelson Repenning »

Contrary to some earlier statements, there is ample evidence that SD can be
used to model exactly the issues referred to below. All of this comes,
however, with one important caveat that I think is the source of confusion:

Paraphasing an early statement by Professor Forrester, SD can be used to
model anything that changes over time. Thus, taking one of the example
raised in the previous message, SD would not be particularly useful in
helping guide team selection *IF* you believe that team performance is
totally determined by unchanging attitudinal and dispositional factors of
the team members. If this were the case, then you probably would just use
one of the many personality assessment tools to select the best mix of team
members. However, there is ample evidence that this is not the case, team
performance often has a endogenous component--the nature of the
interactions between members evolves over time. Thus, initial attitudes
and dispositions play a role, but they are not the whole story; one also
needs to understand the dynamics that such a system might generate. SD is
useless for performing the initial assessment of the various personality
traits, but is ideally suited for tackling the problem of understanding the
subsequent evolution.

The SD literature contains numerous examples of models used to understand
the evolution of human behavior, strategic or otherwise. there are a
number of examples in Jay Forresters early work. More recently my own
work often tackles such interpersonal and behavioral dynamics. For
example, in our study of Analog Devices, John Sterman and I modeled how
employee commitment to participating in TQM evolved based on the obsevation
of the actions of management such as goal setting, the allocation of
support resources and willingness to maintain job security. The paper and
technical documentation of this model are available at:

http://web.mit.edu/jsterman/www/ADI/ADIhome.html

also see:

http://web.mit.edu/nelsonr/www/simulation._v3.html


John and I have also used SD to understand the structural causes of discord
between various groups in large organizations such as machine operators and
plant managers and engineers and project leads. See:

http://web.mit.edu/nelsonr/www/SelfConf ... _v1.0.html


In all of these cases we find that initial personality factors are far less
important than the resulting dynamics in determining the observed outcome.
Thus, as is often the case, the high leverage point for improving
performance does not lie in selecting the right people but in designing a
better system.

Hope this helps,

nelson

--------------------------------------------------------------
Nelson P. Repenning
Assistant Professor of Management
Operations Management/System Dynamics Group
Alfred P. Sloan School of Management, MIT
30 Wadsworth St, E53-339
Cambridge, MA 02139
phone: 617258-6889 fax:617258-7579 e-mail:nelsonr@mit.edu
http://web.mit.edu/nelsonr/www
Locked