Elements of a computational infrastructure for social simulation.

Abstract

Applications of simulation modelling in social science domains are varied and increasingly widespread. The effective deployment of simulation models depends on access to diverse datasets, the use of analysis capabilities, the ability to visualize model outcomes and to capture, share and re-use simulations as evidence in research and policy-making. We describe three applications of e-social science that promote social simulation modelling, data management and visualization. An example is outlined in which the three components are brought together in a transport planning context. We discuss opportunities and benefits for the combination of these and other components into an e-infrastructure for social simulation and review recent progress towards the establishment of such an infrastructure.

Full text

Page 1

Computational Infrastructure for Social SimulationPage 1 of 3
The Elements of a Computational Infrastructure for Social Simulation
Mark Birkin1, Rob Allan2, Sean Beckhofer3, Iain Buchan4, June Finch5, Carole Goble3, Andy Hudson-Smith6, Paul
Lambert7, Rob Procter5, David de Roure8, Richard Sinnott9
[1] School of Geography, University of Leeds; m.h.birkin@leeds.ac.uk (corresponding author) [2] STFC, Dares-
bury [3] School of Computer Science, University of Manchester [4] School of Medicine, University of Man-
chester [5] School of Social Sciences, University of Manchester [6] Centre for Applied Spatial Analysis, UCL
[7] Applied Social Science, University of Stirling [8] Electronics and Computer Science, University of South-
ampton [9] NeSC, University of Glasgow
1Introduction
Social simulation is a relatively new and rapidly
expanding field with applications across the aca-
demic research community. Both micro-simulation
and agent-based modelling have been widely
adopted within economics, sociology and geogra-
phy. Simulation models have also provoked high
levels of interest in healthcare research, anthropol-
ogy and political science. Policy interest is substan-
tial at local and national government levels in a
range of forecasting scenarios, including transport,
housing, education and healthcare, and within the
private sector.
The authors of this paper are leading efforts to con-
struct a National e-Infrastructure for Social Simula-
tion (NeISS) in the UK. This paper will present a
review of the objectives and plans of the NeISS in-
frastructure, and outline current technical ap-
proaches and challenges. This £1.5M project, which
is funded under the JISC Information Environment
programme, is both important and timely for a num-
ber of reasons: there is a growing demand for social
simulation ‘services’ among the UK academic
community, and in the public and private sectors;
while the UK National Centre for e-Social Science
(NCeSS) offers a increasingly substantial critical
mass of expertise. Some key NCeSS projects which
are partners in NeISS are reviewed in Section 2 of
this abstract. The dissemination of tools and ser-
vices is backed through a network of international
collaborations involving both European and Asian
Grid Infrastructures; and solid foundations have
been laid for this ambitious programme, which we
refer to in Section 3 of the abstract. There is de-
monstrable user engagement which has been stimu-
lated by ongoing innovations in e-social simulation,
while buy-in from the research community is as-
sured through the prospect of return on existing
investments in both computational and data infra-
structures.
2 Objectives of the NeISS Infrastructure
The MoSeS NCeSS Node at the University of Leeds
(Models and Simulation in e-Social science) is the
precursor to NeISS [1]. MoSeS is concerned with
the projection of simulated city and regional popula-
tions into the medium-term future, from 2006 until
2031. There are no existing projections which sup-
port planning the delivery of services like housing,
healthcare and transportation at the intra-urban
scale.
Consider the example in Figure 1. A category of co-
dependent households has been identified which are
of critical importance for the delivery of social care
(co-dependent households are elderly couples living
together in which one household member is heavily
dependent upon the other). Through detailed simu-
lation modelling the current and future distribution
of such households has been determined.
The model has three components: a population re-
construction method; a dynamic forecasting model;
and a scenario-based (what if?) impact analysis tool
with social and behavioural modelling functionality.
MoSeS social care simulations combine data from
the ONS mid-year population estimates and sub-
national population projections, census data,

Page 2

Computational Infrastructure for Social SimulationPage 2 of 3
Figure 1. Simulation of co-dependency in Leeds
Census output is Crown copyright and is repro-
duced with the permission of the Controller of
HMSO and the Queen’s Printer for Scotland.
2001 Census, Output Area Boundaries. Crown
copyright 2003.
longitudinal micro-data from the British Household
Panel Survey and third party information from
Leeds City Council Social Services and the Leeds
Teaching Hospitals Trust.
NeISS also builds on contributions from three other
NCeSS nodes: GeoVUE, DAMES, and Obesity e-
Lab. GeoVUE is concerned with the provision of
extended visualisation capabilities for analysis and
validationof simulation
Web2.0 mapping tools, and building on existing
work which encourages crowd-sourcing of data [2].
The DAMES Node focuses upon the support and
documentation of activities in manipulating and pre-
paring data for analysis. It's input to the NeISS in-
frastructure will involve embedding tools for docu-
mentation, standardisation and enhancement of data
management activities within the simulation re-
source. This will provide environments to allow
researchers not simply to supply their data to the
simulation service, but also to perform reviews and
manipulations of that data in a manner which is con-
sistent with social science data standards.
e-Lab is a multi-disciplinary team operating across
Health, Social and Computer Sciences. This activity
is heavily application focused, enabling social and
biomedical researchers to share data, information
and analytical tools for obesity research.
model outputs using
Obesity
The contributions of the various NCeSS Nodes de-
scribed above are combined flexibly to produce so-
cial simulation workflows which include primary
datasets, simulation code, tools for analysis and
visualisation of scenario outputs and archives for the
deposition of results from simulation experiments or
scenarios (see Figure 2).
computational means for the composition and en-
actment of such workflows will be discussed.
In the next section, the
Figure 2. Example social simulation workflows
3Computational Requirements
Ne-ISS consists of four interacting layers (see Fig-
ure 3). The first is a service layer comprising the
fundamental components on which simulation de-
pends, and which have been described in Section 2
above. The second is a composition layer in which
individual services are composed into workflows
and then published as coherent entities. The integra-
tion of Taverna functionality into a Portal [3] is
planned to provide JSR 168/286 compliant compo-
nents that provide access to enactment, manage-
ment, monitoring and creation of workflows. This
provides the means by which social simulation ser-
vices can be combined into research-oriented ‘ac-
tivities’ as well as providing a means for the intro-
duction of new functionality.
The publishing strand draws on myExperiment [4], a
VRE for the social curation and sharing of scientific
Research Objects, especially workflows and in silico
experiments, facilitating their integration into a por-
tal. An important focus is the incorporation of social
networking functionality to provide extended archiv-
ing and documentation capability for social simula-
tions. Research Objects, aggregations of resources
(data sets, analysis methods, workflows, results,
people) that tell a particular story about an investiga-
tion, experiment or process and capture key infor-
mation about the lifecycle of the investigation (for
example, provenance information about analyses),
facilitating re-use of results and repeatability of ex-
periments. In combination, layers one and two pro-
vide the basis for the integrated support of the ‘so-
cial simulation research lifecycle’ in which intelli-
gence is generated from data, published, shared,
analysed, archived and perpetually recycled.

Page 3

Computational Infrastructure for Social SimulationPage 3 of 3
Above the composition layer sits an architecture
layer. This provides tools and methods needed to
provide portal access to simulation services and
workflows, and to combine these into domain-
specific exemplars. NCeSS has an existing Sakai
community portal. Alternative technologies are also
under consideration.The portal can be accessed
through Internet2 Shibboleth technologies and UK
access management federation. Tools will be pro-
vided to improve overall security and personalising
of content based upon user privileges. Sakai is a VO
management system, not just a portal and role-based
access is part of its framework. It has fine-grained
permissions for each tool based on a user’s role
within a realm and core services provide this infor-
mation to Sakai tools. The issue is how to get this
information to and from external services, e.g., using
SAML. NeISS will be built around a Service Ori-
ented Architecture into which core services and ex-
ternal services are designed to be pluggable, using
frameworks and standards developed under the e-
Framework for Education and Research and OMII-
UK. Core services including data and computation
will be supported through existing e-infrastructures,
including resources hosted by the project partners at
STFC Daresbury, White Rose Grid, and North West
Grid.
Figure 3. Component layers in NeISS
In the deployment layer, Manchester’s e-Research
Centre is coordinating a range of activities to facili-
tate delivery of applications and increase take-up
within both the social science community and ap-
plied users. The social simulation providers intro-
duced in Section 2 have an established foothold in
housing, transport, health and social care, education
and demographic planning. Engagement with inex-
perienced users accessing services which are inte-
grated within domain-specific service users will
provide a clear demonstration of the impact of this
work. However the success of the project will also
be conditioned by its ability to engage with experi-
enced users accessing capabilities at the level of an
individual service or workflow; and expert users
with the capability to introduce their own services,
and build new workflows through combinations of
new or existing services. In addition to their practi-
cal value, exemplars will be used as a means to fo-
cus component services into more meaningful work-
flows, research objects and portlets which address
particular social science research challenges. The
exemplars will draw on dependencies between all of
the other e-infrastructure components and therefore
serve as a focus for the integration of development
activities between all groups.
4 Conclusions
NeISS is a unique project aiming to build simulation
services for social scientists and policy-makers. The
project is characterised by an interest in the deploy-
ment of simulation tools for data integration, model-
ling, analysis and simulation. These tools are ma-
nipulated in an Information Environment (portal)
which facilitates other activities in the research life-
cycle, such as the composition and enactment of
workflows, publication and sharing of analytical
results.
In our presentation to the All Hands conference we
will illustrate the service components of NeISS in
relation to a specific social care application. The
requirements of users, data integration, analytical
functionality and infrastructure components will all
be considered in relation to this specific context.
Acknowledgements
NeISS is funded under the JISC Information Envi-
ronments Programme.
DAMES, GENeSIS and Obesity e-Lab are all
funded as nodes in the ESRC National Centre for e-
Social Science.
References
[1] Townend P, Xu J, Birkin M, Turner A, Wu B
(2009). Modelling and Simulation for e-Social Sci-
ence, Philosophical Transactions of the Royal Soci-
ety A, 367, 2781-2792.
[2] http://www.maptube.org
[3] http://www.taverna.org.uk
[4] http://www.myexperiment.org