Transcript Computer modelling and participatory research

The FIRMA Project is supported by European
Union's Framework 5 Programme for Research
and Development, and by the European
Commission as part of its Key Action on
Sustainable Management and Quality of Water
programme (contract EVK1-CT1999-00016)
Participatory simulations for
developing scenarios in
environmental resource
management
Nigel Gilbert, Sarah Maltby
Tasia Asakawa
University of Surrey
Policy and applied research
• Inform
• Inspire
• Influence
• Develop
• Encourage
Decision-makers
(policymakers)
Communities
2
Academic social science context
• Scepticism about the possibility of
prediction
• Theoretical abstraction important but
application difficult
• Increased demands for relevance and
application
3
A new(ish) approach
• Since the 1960s



Interactive social science
Participatory methods
Action research
• In all these



Stakeholders learn from their peers as well as from social
scientists
Academics are also stakeholders
Praxis
• Tacit as well as formalised knowledge about action and its
consequences
4
Interactive or participatory social
science
• Users and beneficiaries in collaboration
with academics
• Participatory methods have been
advocated as a way of



Empowering the disadvantaged
Involving the powerful
Reducing the distance between academic and
lay discourse
5
Advantages
• Brings different perspectives
• Brings different kinds of knowledge



Lay knowledge
Expert knowledge
Academic knowledge
• Identifies crucial problems
• Stakeholders have some ownership
of results
6
Problems
• Representation of distributed
stakeholders

E.g. ‘the public’
• Dealing with conflict between
stakeholders
• Confidentiality and privacy
• Maintaining the motivation of
participants
7
Agent-based social simulation
• Stakeholders are represented in the model
as agents
• The agents have the goals, beliefs, and
capabilities of the real stakeholders (or
some simplified version of these)
• Then let the model run to see what
happens
• In order to develop scenarios, spot
recurrent patterns of action, identify
unanticipated consequences…
8
But…
• At best, stakeholders can have a ‘God’s eye view’
of the model, observing its outputs, while what
they want is to understand the setting from their
own perspective
• Hence stakeholders either have to do some
translation or (perhaps more likely) they just ignore
the model because the translation is too difficult.
• The model doesn’t give them much help with an
intuitive understanding of the dynamics
9
Putting the user in the model
An alternative is to replace some or even all
of the agents by real stakeholders (or their
representatives)


The model becomes a multi-user strategy
simulation
Analogous to single person vs. multi-player
computer games
10
Advantages
• More engaging for the users
• More realistic

Instead of ‘looking down’ on the model, the
player participates in a virtual setting
• Users can treat the simulation like a flight
simulator


Practice in circumstances that would be
dangerous if carried out in real life
Scenarios can be established in the simulation
as starting points and then users see what
happens from there
11
More advantages
• Conflict between stakeholders can be
observed and/or modelled
• Can provide data for researchers on
what people would do



Elicits tacit knowledge
Not just what they say they would do
And on how they react to others’ actions
that are in response to their actions
(etc.)
12
Distributed multi-user models
• Participants can be anywhere,
provided that they have internet
access

E.g. in their office
• No duration restrictions

Can be involved while doing their
ordinary work
• But


Less motivation without face-to-face
interaction
Technical difficulties less easy to
solve
13
Implementation options

Client-side
• Needs to run on many differently configured PCs
• Java, Javascript
• Inter-player communication hard to implement and
control
• OR

Server side
•
•
•
•
All software runs on a central server
Server generates HTML pages dynamically
Client only needs a standard web browser
Inter-player communication is simple to implement
14
Server side implementation
• Apache web server

Standard web server
• PHP

Scripting language
• All normal programming constructs
• Basic object orientated features
• Good interfaces to other software and libraries
• Relational database


PostgreSQL
MySQL
• TCP/IP or other inter-process communication to other
models
• All this is open source, free and available under the GNU
licence
15
The server
Program
Apache Web
Server
Page
request
PHP
module
HTML
Data
Read/write
Web page
PostgreSQL
database
16
Sample PHP
<?php
PHP:
programming
language
similar to C++
function show_scale($val) {
/* display a bar to show value of $val */
$val=round($val);
if ($val > 10) $val = 10;
if ($val < 0 ) $val = 0;
$colour = ($val >= 5 ? 'grn' : 'red');
echo "<td><img SRC=\"images/bar-$colour-$val.jpg\"
ALT=\"Value=$val\"
width=104 height=14></td>\n";
}
?>
Embedded
HTML
17
Interface between PHP and the
database
$n_msgs = 3;
/* get the last 3 public messages */
$query = new query("SELECT id, sender, recipient,
to_char(timesent, 'HH24:MI on DD Mon')
as senttime,
timeread, msg
FROM msgs
WHERE (recipient = 'All')
ORDER BY timesent DESC LIMIT $n_msgs");
display_msgs($query);
SQL statement
sent to
database
18
The context
• Drought in summer 1976 led to shock to Zurich’s
water supply system



Capacity increased to guarantee a secure supply
But over-supply leads to risk of stagnant water
Water demand has since fallen as a result of water
saving technology and changing business behaviour
• Water utility regarded as inefficient due to high
fixed costs
• Demand management through pricing would allow
parts of the system to be closed

But tariffs ultimately controlled by public through
referenda
19
20
21
22
Playing the game: design choices
• Roles

Stakeholder representatives play their own roles
• They bring their own knowledge to game

Stakeholders play other roles
• Not tied to prior positions and strategies
• Time

Real-time
• Too slow!

Game time
• Player events drive time forwards

Simulated Clock time
23
Computational agents
• The model can include computational
agents as well as ‘real players’ (people)



When real players are absent (on holiday, away
from the office,…)
When real players have not or cannot be
recruited
Test of modelling adequacy: Can they be
distinguished by their actions from real players?
• If all players are agents, game reverts to
being a conventional multi-agent simulation
24
Evaluating the model
• Robustness

Yields policy advice that applies in a range of scenarios
• Transparency

Model is understandable to stakeholders
• Evaluating the process



Is it used?
Is effect lasting?
Has learning occurred?
25
Summary
• Computational models can be used for discovery
or for policy

And possibly for both at the same time
• If they are to be used for policy, their use must be
carefully designed with an understanding of the
policy context
• That context consists of people with many different
pressures, goals, experiences and interests
• And often situations of deep-rooted conflict and
power differences
26
Participatory methods and
simulation
• Multi-agent simulations can profitably be used as a
component of participatory methods, with some
agents being computational and others human
• The design of the simulation will help to recover
and formalise the knowledge of the participants
• The use of the simulation will help to educate the
participants about options and consequences of
action
• The method recognises (as many participatory
methods do not) the inherent conflict in many
settings
27
28