Hi Ken,
Thanks for your message. Please could you explain why being able to predict
how humans behave means that:
<snip>' Things would run well in our big cities and our financial markets
would all work effectively. Iterative gaming simulations would have brought
the predictable victories that commanders expected in the uncontrollable
wars of the past half-century. We would have long ago eradicated some of the
easily managed health problems that arise from problems that are in
principle easy to control through such simple techniques as hand washing and
teaching patients to take their full prescription treatment rather than
stopping when symptoms decrease.'?
I suspect you may be confusing predictability with determinism, and ignoring
the problem of wishful thinking.
I suggest the argument I presented before doesn't need to be substantiated
any more than I did. The evidence presented (the existence of various fields
etc) is sufficient to support the conclusion. Sure it requires thinking
through the reasoning rather than reading citations or referring to whoever
you might consider authorities, but 'self-evident' evidence has the
advantage that you don't need to trust some other codger's reasoning first
(or worse, their discussion) .
Yes, I agree with your claim there is need for a nuanced view.
Predictability is always relative, partial and contingent to need, but
predictability of likely user behaviours doesn't require knowing where every
molecule of the universe is at all times till the next big bang.
There's an argument that its relatively easy to predict the behaviours of
groups of people but less so for individuals. Ok, let's start with
individuals and some coarse level predictions. How about, every individual
is going to die? Too coarse? How about, if an individual is insulted, they
will react? Too coarse still? How about , if an individual is looking at
another individual's face, they will react in a given pattern of emotion?
(well documented) How about, if an individual is used to left to right
reading they will have difficulty with menu processes working right to left
compared to those working left to right? How about if you know someone and
you are given part of a sentence could you infer the likelihood of the next
words? Bit by bit it is possible to fill out the bigger picture of human
predictability with lots of smaller chunks of prediction about elements of
their life
It is effortless to create thought experiments in this realm and from
experience they always point to humans being predictable in some way. In
fact, I'd love to hear of thought experiments or other evidence that show
that humans are NOT predictable in any way.
The wishful thinking basis for predictability assumes somehow people will
operate according to some approved conscious bounded rationality. The
example you gave is good: people don't always take the medicine they are
prescribed. Human motivations do not necessarily align with what other
people wishfully think should be best. Individuals eat too much, don't take
medicines, smoke, drink, argue on the web and a host of other activities
that to an outside observer might not seem obviously predictable. I suggest
that apparent lack of predictability is more a measure of using poor
quality prediction methodics.
As you say, this requires more nuanced discussion.
I feel what we can say at the moment, however, is there are already well
established approaches to accurately predict the behaviours of groups.
Investment in predicting the behaviour of individuals, however, at the
moment appears to have lagged. Having the ability to predict the
behaviours, feelings and thoughts of individuals has lagged except in
particular fields such as psychology/psychotherapy and the criminal justice
system. One possible explanation that seems plausible is there has been less
commercial or political profit in it, and marketing and PR has until
recently focused on mass media. There is now increasing attention to
predicting individual behaviours that aligns with the awareness that
individualised marketing and political persuasion may be more effective, and
the cost -benefit economics are now favourable with reductions in costs of
communication and computing. Some of this individual prediction work is
showing significant accuracy. For example, the prediction of an
individual's future movements on the basis of past mobile records is
reported to be around 93% accurate
(http://www.northeastern.edu/news/stories/2010/02/network_science.html ).
For me, it is of interest where we locate the unit of analysis in devising
prediction methods for individuals and groups. Many have focused the unit of
analysis on behaviour and effects on behaviour. Pentland and Liu (1999)
claimed it was better to analyse humans in terms of control events.
Personally, I suspect it is better at higher levels still.
Best wishes,
Terry
-
Dr Terence Love
M: 0434975848
[log in to unmask]
=
Pentland, A., Liu, A. (1999) Modeling and Prediction of Human Behaviour,
Neural Computation, `11, 229-242
-----Original Message-----
From: PhD-Design - This list is for discussion of PhD studies and related
research in Design [mailto:[log in to unmask]] On Behalf Of Ken
Friedman
Sent: Friday, 6 September 2013 3:25 PM
To: [log in to unmask]
Subject: Predicting Human Behavior
Dear Terry,
While human beings often have "a limited range of responses that they apply
in a very routine fashion," this does not mean that "humans behave very
predictably."
If it were as easy to predict human behavior as you suggest, we'd be living
in a very different world. Things would run well in our big cities and our
financial markets would all work effectively. Iterative gaming simulations
would have brought the predictable victories that commanders expected in the
uncontrollable wars of the past half-century. We would have long ago
eradicated some of the easily managed health problems that arise from
problems that are in principle easy to control through such simple
techniques as hand washing and teaching patients to take their full
prescription treatment rather than stopping when symptoms decrease.
But human beings don't behave as we predict they will. Not even when their
own lives may depend on it, as they do when patients take prescriptions.
The reason I find these kinds of posts so problematic is that you are not
merely playing devil's advocate. You are arguing too strong a position with
too little evidence.
Simulations exist. Algorithms have value. Design theory can do better.
But not with overconfident kind of reliance you propose on computing power.
I'd be curious to see a well-substantiated argument that based on more than
reasoning. You're claiming this works. That requires evidence and a
well-structured argument.
In another thread, I pointed to Bill Starbuck's (2009) article, "The
constant causes of never-ending faddishness in the behavioral and social
sciences." One of the issues to which Starbuck points is that the behavioral
and social sciences do not work as physics and engineering do. You are an
engineer making a claim about predictability in human behavior. The claim of
algorithmic predictability is among the recurring fads in behavioral and
social science that have not yet proven feasible.
In part, I agree with you, but there are significant nuances to what we can
predict, how we can do so, and how we can make our predictions useful. In a
sense, this kind of predictability across a range of potential behavioral
options is the basis of design affordances. In a larger theoretical sense,
however, you are claiming far too much. A more nuanced and careful argument
would make this a far more useful argument. The devil is in the details.
For those who wish to read Starbuck's article, I have posted it on my
Academia.edu site. It will remain there until September 20 under Teaching
Documents. It is at URL:
http://swinburne.academia.edu/KenFriedman
Yours,
Ken
Ken Friedman, PhD, DSc (hc), FDRS | University Distinguished Professor |
Swinburne University of Technology | Melbourne, Australia |
[log in to unmask]<mailto:[log in to unmask]> | Mobile +61 404 830
462 | Home Page
http://www.swinburne.edu.au/design/people/Professor-Ken-Friedman-ID22.html<h
ttp://www.swinburne.edu.au/design> Academia Page
http://swinburne.academia.edu/KenFriedman About Me Page
http://about.me/ken_friedman
Guest Professor | College of Design and Innovation | Tongji University |
Shanghai, China
--
Starbuck, William H. 2009. "The constant causes of never-ending faddishness
in the behavioral and social sciences." Scandinavian Journal of Management.
Vol. 25, No. 1, pp. 108-116.
doi:10.1016/j.scaman.2008.11.005
Available through September 20 under Teaching Documents at this URL:
http://swinburne.academia.edu/KenFriedman
--
Terry Love wrote:
-snip-
The reality is that humans behave very predictably, and have a limited range
of responses that they apply in a very routine fashion.
-snip-
-----------------------------------------------------------------
PhD-Design mailing list <[log in to unmask]> Discussion of PhD
studies and related research in Design Subscribe or Unsubscribe at
https://www.jiscmail.ac.uk/phd-design
-----------------------------------------------------------------
-----------------------------------------------------------------
PhD-Design mailing list <[log in to unmask]>
Discussion of PhD studies and related research in Design
Subscribe or Unsubscribe at https://www.jiscmail.ac.uk/phd-design
-----------------------------------------------------------------
|
