StrategiesForTheFuture
Task 6 of RIPPLE project aims to build an ES to forecast future policies and strategies that could [better] promote particular quality products and services (QPS) in the study regions. Thus whilst empirical studies analyse policies and strategies that are currently used to promote particular QPS in the study regions, the RIPPLE ES aims to forecast future policies and strategies that could contribute to the better promotion of QPS in these study regions.
Because future is plural, and because an overall and complete model of the role played by the factors and actors influencing the creation, development and maintenance of QPS in a given area currently does not exist, the RIPPLE ES aims to point out future institutional structures, public policies, and marketing strategies that could contribute to the better promotion of QPS
in certain lagging areas, by using three methods:
1. by pointing out the main current strategies of producers and institutions (alternative strategies) ; to do this, RIPPLE ES uses two methodologies: Probabilistic Reasoning (Spiegelhalter (1993), Boerlage (1994), Heckerman (1995), Zhang (1996)) and Case-Based Reasoning (Riesbeck and Schank. (1989), Kolodner (1991).
Probabilistic Reasoning
Probabilistic Reasoning Module of RIPPLE ES searches for the most appropriate current strategies/ policies to improve the competitive advantage of a given sector/ region, by referring to the probabilistic relations (Bayesian belief networks (Bbn)) existing between the actions and the achievement of the typical objectives of producers/ institutions, in a given study region / sector (4).
Such a Probabilistic Analysis includes four steps:
a. End-User of ES loads the BBN describing the probabilistic relations existing between the objectives and the actions (and the contextual features of these ones) of the producers/ institutions involved in the production/ promotion of certain QPSs.
b. End-User selects the set of typical objectives to be achieved (ref. Appendix VARS-A1/2)
c. ES points out the most appropriate actions and contextual features to achieve these objectives, by computing the most probable explanation for the achieving of these ones
d. ES points out the sensitivity of each objective to actions and contextual features.
Case-Based Reasoning
Rather than searching for the most appropriate strategies/ policies to improve the comparative advantage of a given sector/ region (ref. Probabilistic Reasoning), Case-Based Reasoning Module of RIPPLE ES searches for the alternative strategies/ policies that certain producers/ institutions could adopt to better achieve their own objectives/ missions. To do this, CBR Module refers to strategies/ policies used by producers/ institutions in similar contexts, and to reasoning rules specifying the ways to be used to find out the case-studies that are 'comparable/ similar' to a given target producer/ institution (matching rules) and to adapt the strategies/ policies of these ones to the target case (adaptation rules).
So, a Case-Based Reasoning Session includes the following steps:
a. End-User of ES loads Data and Reasoning Rules to be used by ES (5)
b. End-User describes the structural and contextual features of a target producer/ institution X by referring to appropriate criteria (ref. Appendix VARS-A1/2-B1/2-/C1/2)
c. ES searches for the studied producers/ institutions that match to the target case X, by referring to appropriate matching rules and assessment criteria (ref. Append. CBR: RULES A)
d. End-User selects one of the L most effective retrieved cases, and asks ES to adapt aims, targets, actions of this one to case X, via adaptation rules (ref. App. CBR: RULES B)
e. End-User stores the alternative strategy/ policy X[y] pointed out by ES, or searches for an other one by referring to an other effective retrieved case.
------------------
(4) these probabilistic relations (Bbn) have been built by Cemagref from data collected via producer and institutional surveys
(5) these Data and Rules have been built by Cemagref from data collected via producer/ institution surveys
2. by assessing and ranking these strategies, by referring to their internal and regional impacts : to do this, RIPPLE ES uses Probabilistic Analysis ( Propagation of Findings), Consensus Analysis (Changes in Alignements), and Actor-Network Analysis (Direct Relations and Joint Involvements). Ranking strategies is achieved by using Multicriteria Analysis and Sensitivity Analysis.
RIPPLE ES can aid to assess the regional implications of the I {I=K*L} possible alternative strategies/ policies pointed out by the CBR Sessions (6), by computing the consequences of these ones on the following features:
(i) the competitive advantages of a given sector/ region
(ii) the perceptions of the current products, regional images and quality signs by consumers.
(iii) the alignments of/ the consensus among producers and/or institutions
(iv) the interactions among institutions
To do this, RIPPLE ES uses three methodologies: Probabilistic Reasoning, Consensus Analysis (Romney & al (1987)), and Network Analysis (Mizruchi & Galaskiewicz (1994).
Probabilistic Reasoning
Firstly, Probabilistic Reasoning Module of RIPPLE ES can aid to assess the regional implications of possible alternative strategies/ policies, by computing the consequences of each of these "findings/evidences" on the competitive advantage of a given sector/ region, by referring to the probabilistic relations existing among the typical actions and objectives of producers/ institutions of this study region / sector.
So, such a Probabilistic Analysis includes the following steps:
1. End-User of ES loads the Bayesian Belief Network (Bbn) describing the probabilistic relations between actions and objectives in a given sector/ region
2. For each possible strategy/ policy X[y]:
. End-User inputs in the Beliefs Screen, as new findings/evidences, the actions, targets and beliefs that actor X could adopt
. ES computes the consequences of this "finding/evidence" on the probability of achievement of each typical objective of producers/ institutions, by updating the Bbn.
Secondly, Probabilistic Reasoning Module of RIPPLE ES can aid to assess the regional implications of possible alternative strategies/ policies pointed out by the CBR Sessions, by computing the consequences of each of these "findings/evidences" on the perceptions of the current products, regional features/images and quality signs by consumers, by referring to the probabilistic relations existing between certain actions/ beliefs and certain behavioural features of the consumers.
This Probabilistic Analysis includes the following steps:
1. End-User of ES loads the Bbn describing the probabilistic relations existing between certain marketing strategies/ policies and the perceptions of consumers
2. For each consumers' behaviour :
- End-User select inputs the finding he/she would study (ref. Appendix VARS-E)
- ES points out the most probable explanation (actions and contextual features) of this behaviour, and computes the sensitivity of this one
3. For each possible alternative stategy/ policy X[y]:
- End-User translates this one in terms perceived by consumers (ref. VARS-B1/2)
- End-User inputs these actions/ perceptions as new findings/evidences in Beliefs Screen
- ES computes the consequences of these " findings/evidences " on the probability of each behavioural feature of consumers
- End-User analyses the probabilistic relations existing between perceptions, behaviours and social features of consumers via the Beliefs Screen.
Consensus Analysis
Consensus Analysis (CA) Module of RIPPLE ES aids to assess regional implications of alternative strategies/ policies, by computing the consequences of each of these ones: (i) on the consens degrees of actions (VARS-B1/2) and positions (ref. VARS-G1/2) used by the producers/ institutions involved in the production or promotion of certain QPSs in a given region, and (ii) on the causes of divergence (actions, positions and actors) among these producers and institutions.
A Consensus Analysis includes the following steps:
1. End-User of ES loads the Matrix M describing the current objectives, actions, targets (ref. App. VARS A&B1/2) or the system of values (ref. App. VARS G1/2) of the producers and institutions involved in the production or promotion of certain QPSs in a given region.
2. ES computes the current consensual actions and positions of producers and/or institutions, as well as the main causes of divergence among these ones
3. For each possible alternative strategy/ policy X[y]:
- End-User inputs in Matrix M, the changes in actions/ system of values corresponding to this strategy/ policy, (ref. Appendix CA: Example of Matrix M)
- ES computes future consensual actions and positions (and causes of divergence)
- End-User analyses differences between current and future situations
Network Analysis
Network Analysis (NA) Module of RIPPLE ES aids to assess regional implications of policies pointed out by CBR Sessions, by computing the consequences of each of these ones on:
(i) the cohesion of the institutional networks
(ii) the equivalent institutional positions and roles in these networks,
(iii) the prominence (power) of certain institutional positions and roles in these networks, as well as the opportunities for certain institutions to take advantage of competition among them.
In summary, Network Analysis aims at pointing out the consequences of alternative policies on the current institutional boundaries, bonds (underlying shared beliefs) and behaviours.
A Network Analysis includes the following steps:
1. End-User of ES loads the Adjency Matrix N describing the N*N institutional interactions to be analysed (ref. App. MAPS)
2. ES display the graph of relations among institutions corresponding to the Adjency Matrix N
3. ES computes the current values of the following features (ref. NA: Examples):
(i) the cohesion distances, distances by dyad, density table, cliques/clans...
(ii) the density(7), equivalent distances, triads patterns, clusters (8)
(iii) the powers, centrality(9), equilibrium relations by dyad
(iv) the ego and dyads indices, hole signature, constraints relations (10)
4. For each possible alternative strategy/ policy X[y]:
- End-User inputs the changes in induced by this one in the Adjency Matrix N
- ES computes the future value of the four above features (ref. 3 i, ...iv)
- End-User analyses differences between current and future situations
-------------------
(6) or coming from any assumption about changes in structural characteristics, policies/strategies, and system of values of the modelled actors, that the End-User would test (ref. TABLE D, lines 59, 60, 61 and TABLE2 Q17a....)
(7) density is used to measure levels of uncentralized inter-organisational coodination (Scott (1991)),
(8) clusters are used to identify sets of actors with similar relations/roles in the network (Mizruchi & Galaskiewicz (1994))
(9) closeness centrality and betweenesss centrality are commonly used to measure the degree of independence of a given actor from others in a network; global centrality measures the degree to which an entire network is focused around a few central actors (Scott (1991)), i.e. its the degree to which relations are guided by the formal hierarchy, the degree to which an inter-organisational network is dominated by a few places (Irwin and Huges (1992)).
(10) the network constraints in a relation defines the cost of negociation in the relation ; the hole signature summarise the situation
Finally, RIPPLE ES can aid to point out ways to be used to better promote particular Quality Products and Services in the future, by ranking the assessed alternative strategies/ policies under different assumptions about the relative weight of their internal and regional effects.
To do this, RIPPLE ES combines Multicriteria Analysis (Roy & Vandepooten (1995) and Sensitivity Analysis.
Multicriteria Analysis aims at ranking alternative strategies/ policies, by referring both to their internal effectiveness and to their regional implications. Knowing how it is very difficult to identify the correct weight for each internal objective and regional implication - notably because these ones depend on the [political] point of view of decision makers, Sensitivity Analysis provides a way to compute the variations in the preference indexes induced by the variations of the relative weights of each criteria.
So, a Multicriteria Analysis includes the following stages:
1. End-User of ES loads the Impact Matrix O describing the impact of each alternative strategy/ policy on each typical objective of producers/ institutions (ref. VARS A1/2) and each regional implication (ref. App. VARS H1/2)
2. ES computes the preference index of each strategy/ policy from this matrix, by giving the same weight to each internal objective and regional implication.
3. End-User of ES sets the relative weight of each (or sets of) criteria by referring to the points of view of a given decision maker (ref. system of values).
4. ES computes and shows in a graphical format, the best and the worst actions. As well as the 'behaviour' of each strategy/ policy (i.e. the values of the preference index) for each distribution of weights.
3. by computing the regional implications of the dissemination of the best strategies as scenarios. Scenarios are not predictions. Rather, by offering a basis for discussing the consequences of some possible future strategies/ policies, they would be communication devices to bound the uncertainty of the future (Brewer 1983, Shoemaker 1989, Godet 1991, Schwartz 1993). By making more transparent both positive and negative ripple effects of the dissemination of certain strategies, the scenarios generate some focus of attention that could have implications in the construction of future institutional policies and structures. They can be viewed as ex ante cost-benefit analyses, which would support the decisions of public institions.