Synergies and conflicts on the Synergies and conflicts on the - - PowerPoint PPT Presentation

Synergies and conflicts on the Synergies and conflicts on the landscape of domestic energy landscape of domestic energy consumption: beyond metaphor consumption: beyond metaphor

Alternate title Alternate title

“ “What the heck is a What the heck is a ‘ ‘Bayesian Belief Network Bayesian Belief Network’ ’? ?” ”

David Shipworth David Shipworth University of Reading University of Reading

Overview Overview

Paper contains three interrelated elements: Paper contains three interrelated elements:

1. 1.

Theoretical: Theoretical:

Operationalising Operationalising the concept of the concept of ‘ ‘barriers barriers’ ’ through linking through linking ‘ ‘socio socio-

• technical landscapes

technical landscapes’ ’ to to complex systems theory and complex systems theory and BBNs BBNs. .

2. 2.

Methodological: Methodological:

Why Bayesian Belief Networks may be useful Why Bayesian Belief Networks may be useful for modelling home energy use. for modelling home energy use.

3. 3.

Method: Method:

How to construct a Bayesian Belief Network How to construct a Bayesian Belief Network

Context Context

• UK

UK ‘ ‘Carbon Vision Carbon Vision – – Buildings Buildings’ ’ [2005 [2005-

• 2008]

2008]

• £

£5M from Carbon Trust; EPSRC; ESRC 5M from Carbon Trust; EPSRC; ESRC

• Three Consortia:

Three Consortia:

• BMT

BMT ‘ ‘Building Market Transformation Building Market Transformation’ ’ ~ ~£ £1M 1M

• TARBASE

TARBASE ‘ ‘Transforming the Asset Base Transforming the Asset Base’ ’ ~ ~£ £1M 1M

• CaRB:

CaRB: ‘ ‘Carbon Reduction in Buildings Carbon Reduction in Buildings’ ’ ~ ~ £ £3M 3M

• CaRB: Non

CaRB: Non-

• domestic;

domestic;

• CaRB: Longitudinal

CaRB: Longitudinal

• CaRB: Community

CaRB: Community

• CaRB: Socio

CaRB: Socio-

• Technical (~

Technical (~£ £0.8M) 0.8M)

• University of Reading & University of Newcastle

University of Reading & University of Newcastle

Research Question Research Question

• Same house + differing occupants =

Same house + differing occupants = radically differing energy use. Why? radically differing energy use. Why?

Energy Use

Differing perspectives, Differing perspectives, epistemologies and data epistemologies and data-

• types

types

• Perspectives

Perspectives

• Architectural (Passive solar design

Architectural (Passive solar design… …) )

• Engineering (Envelope & services

Engineering (Envelope & services… …) )

• Economic (Rational action model

Economic (Rational action model… …) )

• Psychological (Attitude

Psychological (Attitude-

• Behaviour

Behaviour… …) )

• Sociological (Context & contingency

Sociological (Context & contingency… …) )

• Epistemologies

Epistemologies

• From na

From naï ïve realism to utility maximisation to social ve realism to utility maximisation to social constructivism constructivism

• Need

Need ‘ ‘common ground common ground’ ’/ /‘ ‘common language common language’ ’

• Data types

Data types

• qualitative, quantitative, discrete, continuous

qualitative, quantitative, discrete, continuous

Complex interrelated factors Complex interrelated factors

• Household size; Income; Floor area;

Household size; Income; Floor area;

• Income; Education level; Gender

Income; Education level; Gender

• Environmental

Environmental ‘ ‘attitudes attitudes’ ’; Energy practices ; Energy practices

• Market segmentation differences

Market segmentation differences

• Energy

Energy ‘ ‘visibility visibility’ ’; Specific knowledge ; Specific knowledge

• Grandmother

Grandmother’ ’s shoe size? s shoe size?

Mediating & Conditioning variables Mediating & Conditioning variables

• Education

Education is + is +ve ve correlated with correlated with income income. .

• Income

Income is + is +ve ve correlated with correlated with product ownership product ownership (mediating variable) (mediating variable)

• Product ownership

Product ownership is + is +ve ve correlated with correlated with Home Home energy use energy use

• Education

Education is + is +ve ve correlated with correlated with Environmental Environmental awareness awareness

• Awareness

Awareness is is – –ve ve correlated with correlated with energy use energy use in in some some market segments market segments (conditioning variable) (conditioning variable)

Representation ( Representation (‘ ‘Netica Netica’ ’) )

+ve +ve +ve +ve

• ve but strength conditional on MS

Bayesian Belief Networks Bayesian Belief Networks

• Integration of qualitative and quantitative data from

Integration of qualitative and quantitative data from experts, case studies, data experts, case studies, data-

• sets and models;

sets and models;

• Integrate of new data as it becomes available;

Integrate of new data as it becomes available;

• Use of categorical and continuous variables;

Use of categorical and continuous variables;

• Highlight conflicts or synergies between variables.

Highlight conflicts or synergies between variables.

• Intuitive display of relationships between variables;

Intuitive display of relationships between variables;

• Straightforward sensitivity testing.

Straightforward sensitivity testing.

• ‘

‘Subjective probability Subjective probability’ ’ provides common epistemological provides common epistemological ‘ ‘common ground common ground’ ’ between social and engineering between social and engineering approaches approaches

• Create consensus based decision support systems;

Create consensus based decision support systems;

Network Construction: Phase 1 Network Construction: Phase 1

• A Realist Review of the literature

A Realist Review of the literature

• Stakeholder engaged

Stakeholder engaged

• Stresses context and contingency

Stresses context and contingency

• Tests

Tests ‘ ‘theories theories’ ’ against the evidence base against the evidence base

• Data extraction templates tailored to BBN.

Data extraction templates tailored to BBN.

• Focus on factors influencing home energy

Focus on factors influencing home energy use, their strength, direction and evidence use, their strength, direction and evidence base base

Network Construction: Phase 2 Network Construction: Phase 2

1. 1.

Sensitivity analysis Sensitivity analysis: Which variables are key? : Which variables are key?

2. 2.

Uncertainty analysis Uncertainty analysis: Does the network as a : Does the network as a whole remain within expected bounds. whole remain within expected bounds.

3. 3.

Pruning Pruning: Delete insensitive variables and links. : Delete insensitive variables and links.

4. 4.

Refining Refining: Additional quantitative analysis to : Additional quantitative analysis to reassess key probabilities. reassess key probabilities.

5. 5.

Extending Extending: :

• Primary qualitative research to extend the network

Primary qualitative research to extend the network and refine contingencies and refine contingencies

• Primary quantitative research is conducted to

Primary quantitative research is conducted to populate new nodes with probability data. populate new nodes with probability data.

6. 6.

Go to step 1(repeat Go to step 1(repeat ‘ ‘till money or time runs out!) till money or time runs out!)

Representation Representation

+ve +ve +ve +ve

• ve but strength conditional on MS

‘Frustrated’ pathways (takeback)

Fitness landscape Fitness landscape

http:// http://www.talkorigins.org/design/faqs/nfl/sm.jpg www.talkorigins.org/design/faqs/nfl/sm.jpg

Socio Socio-

• technical landscape

technical landscape

( (Sahal Sahal, 1985: p.79 cited in , 1985: p.79 cited in Geels Geels 2001) 2001)

Conclusions Conclusions

• Provides

Provides policy focused decision support policy focused decision support

• Supports evidence based policy making

Supports evidence based policy making

• Transdisciplinary research epistemology

Transdisciplinary research epistemology

• K

Knowledge synthesis nowledge synthesis consistent with Realist consistent with Realist Review method Review method

• Models

Models ‘ ‘learn learn’ ’ through through continual integration of data continual integration of data

• Provides

Provides ‘ ‘cross cross-

• fertilisation

fertilisation’ ’ between fields between fields

• Models specific

Models specific ‘ ‘take take-

• back

back’ ’ effects effects

• Allows for identification of very specific

Allows for identification of very specific ‘ ‘barriers barriers’ ’ and programme interventions to rectify them. and programme interventions to rectify them.