Abstract
Current decision support systems address domains that are heterogeneous in nature and becoming progressively larger. Such systems often require the input of expert judgement about a variety of different fields and an intensive computational power to produce the scores necessary to rank the available policies. Recently, integrating decision support systems have been introduced to enable a formal Bayesian multi-agent decision analysis to be distributed and consequently efficient. In such systems, where different panels of experts independently oversee disjoint but correlated vectors of variables, each expert group needs to deliver only certain summaries of the variables under their jurisdiction, derived from a conditional independence structure common to all panels, to properly derive an overall score for the available policies. Here we present an algebraic approach that makes this methodology feasible for a wide range of modelling contexts and that enables us to identify the summaries needed for such a combination of judgements. We are also able to demonstrate that coherence, in a sense we formalize here, is still guaranteed when panels only share a partial specification of their model with other panel members. We illustrate this algebraic approach by applying it to a specific class of Bayesian networks and demonstrate how we can use it to derive closed form formulae for the computations of the joint moments of variables that determine the score of different policies.
Similar content being viewed by others
Notes
The CK-class is similar in nature to the SupraBayesian approach in standard Bayesian combination of subjective distributions approach (French 2011). Here we use the terminology CK-class to emphasize that only some specific information needs to be shared and agreed upon by the different panels of experts.
For simplicity, we assume the intercept to be equal to zero since utilities are unique up to positive affine transformations.
We think of \(\theta _{0i}'\) as a parameter although this consists of the sum of a parameter \(\theta _{01}\) and an error \(\varepsilon _i\). Note, however, that from a Bayesian viewpoint these are both random variables.
Notice that these values are then normalized to give utility functions between 0 and 1.
In the multilinear case, higher moments are required. Here we assume that these can be computed from the first two moments using the recursions of normal distributions.
References
Barons MJ, Smith JQ, Leonelli M (2015) Decision focused inference on networked probabilistic systems: with applications to food security. In: Proceedings of the joint statistical meeting, pp 3220–3233
Barons MJ, Wright SK, Smith JQ (2017) Eliciting probabilistic judgements for integrating decision support systems. Technical report
Blangiardo M, Cameletti M (2015) Spatial and spatio-temporal Bayesian models with R-INLA. Wiley, Chichester
Bowen NK, Guo S (2011) Structural equation modeling. Oxford University Press, Oxford
Brandherm B, Jameson A (2004) An extension of the differential approach for Bayesian network inference to dynamic Bayesian networks. Int J Intell Syst 19(8):727–748
Brillinger DR (1969) The calculation of cumulants via conditioning. Ann Inst Stat Math 21:215–218
Castillo E, Gutiérrez J, Hadi A, Solares C (1997) Symbolic propagation and sensitivity analysis in Gaussian Bayesian networks with application to damage assessment. Artif Intell Eng 11(2):173–181
Cooper G, Yoo C (1999) Causal discovery from a mixture of experimental and observational data. In: Proceedings of the 15th conference on uncertainty in artificial intelligence, pp 116–125
Cowell RG, Dawid AP, Lauritzen SL, Spiegelhalter DJ (1999) Probabilistic networks and expert systems. Springer, New York
Cox DA, Little J, O’Shea D (2007) Ideals, varieties and algorithms. Springer, New York
Darwiche A (2003) A differential approach to inference in Bayesian networks. J ACM 50(3):280–305
Dawid AP (1979) Conditional independence in statistical theory. J R Stat Soc B 41:1–31
Dowler E, Lambie-Mumford H (2015) How can households eat in austerity? Challenges for social policy. Soc Policy Soc 14:417–428
Drewnowski A, Specter SE (2004) Poverty and obesity: the role of energy density and energy costs. Am J Clin Nutr 79:6–16
Efendigil T, Önüt S, Kahraman C (2009) A decision support system for demand forecasting with artificial neural networks and neuro-fuzzy models: a comparative analysis. Expert Syst Appl 36:6697–6707
FAO (1996) Rome declaration on world food security and world food summit plan of action. World Food Summit
Faria AE, Smith JQ (1997) Conditionally externally Bayesian pooling operators in chain graphs. Ann Stat 25:1740–1761
Farr AC, Mengersen K, Ruggeri F, Simpson D, Wu P, Yarlagadda P (2019) Combining opinions for use in Bayesian networks: a measurement error approach. Int Stat Rev. https://doi.org/10.1111/insr.12350
Freeman G, Smith JQ (2011) Bayesian MAP model selection of chain event graphs. J Multivar Anal 102:1152–1165
French S (1997) Uncertainty modelling, data assimilation and decision support for management of off-site nuclear emergencies. Radiat Prot Dosim 73(1–4):11–15
French S (2011) Aggregating expert judgement. Rev Real Acad Ciencias Exactas Fisicas Nat Ser A Mat 105(1):181–206
French S, Maule J, Papamichail KN (2009) Decision behaviour, analysis and support. Cambridge University Press, Cambridge
Görgen C, Leonelli M, Smith J (2015) A differential approach for staged trees. In: European conference on symbolic and quantitative approaches to reasoning and uncertainty. Springer, Berlin, pp 346–355
Hernandez T, Bennison D (2000) The art and science of retail location decisions. Int J Retail Distrib Manag 28:357–367
Jensen FV, Nielsen TD (2013) Probabilistic decision graphs for optimization under uncertainty. Ann Oper Res 204:223–248
Johnson S, Mengersen K (2012) Integrated Bayesian network framework for modeling complex ecological issues. Integr Environ Assess Manag 8(3):480–490
Keeney RL, Raiffa H (1976) Decisions with multiple objectives: preferences and value trade-offs. Cambridge University Press, Cambridge
Kennedy MC, O’Hagan A (2001) Bayesian calibration of computer models. J R Stat Soc B 63(3):425–464
Kitchen S, Tanner E, Brown V, Colin P, Crawford C, Deardon L, Greaves E, Purdon S (2013) Evaluation of the free school meals pilot: impact report. Technical report, Department for Education, DFERR227
Koller D, Pfeffer A (1997) Object-oriented Bayesian networks. In: Proceedings of the 13th conference on uncertainty in artificial intelligence, pp 302–313
Lauritzen SL (1992) Propagation of probabilities, means and variances in mixed graphical association models. J Am Stat Assoc 87:1098–1108
Leonelli M, Smith JQ (2013) Using graphical models and multi-attribute utility theory for probabilistic uncertainty handling in large systems, with application to the nuclear emergency management. In: Proceedings of ICDEW, pp 181–192
Leonelli M, Smith JQ (2015) Bayesian decision support for complex systems with many distributed experts. Ann Oper Res 235:517–542
Leonelli M, Smith JQ (2017) Directed expected utility networks. Decis Anal 17(2):108–125
Leonelli M, Görgen C, Smith J (2017) Sensitivity analysis in multilinear probabilistic models. Inf Sci 411:84–97
Loopstra R, Reeves A, Taylor-Robinson D, Barr B, McKee M, Stuckler D (2015) Austerity, sanctions, and the rise of food banks in the UK. BMJ 350:h1775
Madsen AL, Jensen FV (2005) Solving linear-quadratic conditional Gaussian influence diagrams. Int J Approx Reason 38:263–282
Mahoney S, Laskey K (1996) Network engineering for complex belief networks. In: Proceedings of the 12th international conference on uncertainty in artificial intelligence. Morgan Kaufmann Publishers Inc., New York, pp 389–396
McCullagh P (1987) Tensor methods in statistics. Chapman and Hall, London
Müller SM, Machina MJ (1987) Moment preferences and polynomial utility. Econ Lett 23:349–353
Murphy KP (2002) Dynamic Bayesian networks: representation, inference and learning. PhD thesis. University of California, Berkeley
Nilsson D (2001) The computation of moments of decomposable functions in probabilistic expert systems. In: Proceedings of the 3rd international symposium on adaptive systems, pp 116–121
Pearl J (1988) Probabilistic inference in intelligent systems. Morgan Kaufmann, San Mateo
Pearl J (2000) Causality: models, reasoning and inference. Cambridge University Press, Cambridge
Queen CM, Smith JQ (1993) Multiregression dynamic models. J R Stat Soc B 55(4):849–870
Smith JQ (1994) Plausible Bayesian games. In: Proceedings of the fifth valencia international meeting, 5–9 June 1994, pp 387–406
Smith JQ, Anderson PE (2008) Conditional independence and chain event graphs. Artif Intell 172(1):42–68
Smith JQ, Barons MJ, Leonelli M (2015) Coherent frameworks for statistical inference serving integrating decision support systems. Technical report. arXiv:1507.07394
Spiegelhalter DJ, Lauritzen SL (1990) Sequential updating of conditional probabilities on directed graphical structures. Networks 20:579–605
Sullivant S, Talaska K, Draisma J (2010) Trek separation for Gaussian graphical models. Ann Stat 38:1665–1685
Wall MM, Amemiya Y (2000) Estimation for polynomial structural equation models. J Am Stat Assoc 95(451):929–940
Westland JC (2015) Structural equation modeling: from paths to networks. Springer, New York
Wolfram Research, Inc. (2017) Mathematica, Version 11.1. Champaign
Acknowledgements
We acknowledge that J.Q. Smith was partly supported by EPSRC Grant EP/K039628/1 and The Alan Turing Institute under EPSRC Grant EP/N510129/1, whilst E. Riccomagno was supported by the GNAMPA-INdAM 2017 Project.
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendices
Appendix: Proofs
1.1 Proof of Corollary 1
Adequacy is guaranteed if the EU function can be written in terms of \(\mu _{ji}(d)\) and \(k_{{\varvec{b}}}\), \(i\in [m]\), \(j\in [s_i]\) and \(d\in {\mathcal {D}}\). Note that
The argument of this expectation is a monomial of multi-degree lower or equal to \({\varvec{a}}^*\). Moment independence then implies that \(\bar{u}(d)=\sum _{{\varvec{b}}\in B}k_{{\varvec{b}}}\prod _{i\in [m]}\prod _{j\in [s_i]^0}\mu _{ji}(d),\) and the result follows.
1.2 Proof of Theorem 1
Fix a policy \(d\in {\mathbb {D}}\) and suppress this dependence. Under the assumptions of the theorem, the utility function can be written as
Note also that we can rewrite (16) as
where
and \({\mathcal {P}}^m_0([m])={\mathcal {P}}_0([m])\cap \{m\}\). Calling \({\varvec{\theta }}\) the overall parameter vector of the IDSS, the conditional EU function can be written applying sequentially the tower rule of expectation as
From Eq. (17), the definition of a polynomial SEM and observing that the power of a polynomial is still a polynomial function in the same arguments, it follows that
where \(p_m\) is a generic polynomial function. Thus \(\hat{u}({\varvec{Y}}_{[m-1]})+{\mathbb {E}}_{Y_m|{\varvec{Y}}_{[m-1],{\varvec{\theta }}}}\left( \hat{u}({\varvec{y}}_{m})\right)\) is also a polynomial function in the same arguments. Following the same reasoning, we have that
where \(p_{m-1}\) is a generic polynomial function. Therefore the same procedure can be applied to all the expectations in (18). So \({\mathbb {E}}(u({\varvec{Y}})\;|\;{\varvec{\theta }})=p_1({\varvec{\theta }}),\) where \(p_1\) is a generic polynomial function. This defines by construction an algebraic conditional EU, where the functions \(\lambda _{ij}\) are monomials. Quasi independence and Lemma 1 then guarantee score separability holds.
1.3 Proof of Proposition 2
We prove Eq. (10) via induction over the indices of the variables. Let \(Y_1\) be a root of \({\mathcal {G}}\). Thus \(Y_1=\theta _{01}'\), where \(\theta _{01}'\) is the monomial associated to the only rooted path ending in \(Y_1\), namely \((Y_1)\). Assume the result is true for \(Y_{n-1}\) and consider \(Y_{n}\). By the inductive hypothesis we have that, if \(i<j\) whenever \(i\in \varPi _j\),
Note that every rooted path ending in \(Y_n\) is either \((Y_n)\) or consists of a rooted path ending in \(Y_i\), \(i\in \varPi _n\), together with the edge \((Y_i,Y_n)\). From this observation the result then follows by rearranging the terms in Eq. (19). Equation (11) can be proven via the same inductive process noting that \({\mathbb {E}}(Y_1\;|\;{\varvec{\theta }},d)=\theta _{01}'\) and \({\mathbb {E}}(Y_n\;|\;{\varvec{\theta }},d)=\theta _{0n}'+\sum _{i\in \varPi _n}\theta _{in}{\mathbb {E}}(Y_i\;|\;{\varvec{\theta }},d)\).
1.4 Proof of Theorem 2
Under the assumptions of the theorem, the conditional EU function can be written as in Eq. (12). From the linearity of the expectation operator we have that
Applying moment independence and letting \(V_i\) and \(E_i\) be the sets of distinct vertices and edges, respectively, for all the elements \(P\in {\mathbb {P}}_i\), we have that
where \(c_{ik}\) is the element of \({\varvec{c}}_i\) associated to \(\theta _{jk}\) and \(Ch_l\) is the index of a children of the vertex l. The thesis then follows since each of these expectations is delivered by an individual panel.
1.5 Proof of Lemma 3
To prove this result we first show that under the assumptions of the lemma the utility function can be written as
and then prove that
The lemma then follows by substituting into Eq. (20) for \({\varvec{y}}^{{\varvec{a}}}\) given in Eq. (21).
Fix a policy \(d\in {\mathbb {D}}\) and suppress this dependence. We prove Eq. (20) via induction over the number of vertices of the BN. If the BN has only one vertex then
This can be seen as an instance of Eq. (20). Assume the result holds for a network with \(n-1\) vertices. A multilinear utility factorisation can be rewritten as
The first term on the rhs of (22) is by inductive hypothesis equal to the sum of all the possible monomial of degree \({\varvec{a}}=(a_1,\ldots ,a_{n-1},0)\) where \(0\le a_i\le n_i\), \(i\in [n]\). The other terms only include monomials such that the exponent of \(y_n\) is not zero. Letting \({\varvec{n}}_{n-1}=(n_i)_{i\in [n-1]}\), \({\varvec{y}}_{[n-1]}=\prod _{i\in [n-1]}y_i\) and \(u'=\sum _{I\in {\mathcal {P}}_0^n([n])}k_I\prod _{i\in I{\setminus } \{n\}}u_i(y_i)u_n(y_n)+k_nu_n(y_n)\), we now have that
Therefore, Eq. (20) follows from equations (22) and (23). To prove Eq. (21) note that the monomial \({\varvec{Y}}^{{\varvec{a}}}\) can be written as
Equation (21) then follows by noting that
1.6 Proof of Theorem 3
Under the conditions of the theorem, the conditional EU function can be written as in (15). The linearity of the expectation operator than implies that
Applying moment independence and letting \(V_{tot}\) and \(E_{tot}\) be the sets of distinct vertices and edges, respectively, for all the elements \(P\in {\mathbb {P}}=\cup _{i\in [m]}{\mathbb {P}}_i\), we then have that for any \({\varvec{l}}\simeq {\varvec{b}}\)
Score separability then follows since each of these expectations is delivered by an individual panel.
Code for the multilinear factorization
Rights and permissions
About this article
Cite this article
Leonelli, M., Riccomagno, E. & Smith, J.Q. Coherent combination of probabilistic outputs for group decision making: an algebraic approach. OR Spectrum 42, 499–528 (2020). https://doi.org/10.1007/s00291-020-00588-8
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00291-020-00588-8