Method for quantifying social impact indicators, apparatus for quantifying social impact indicators, and program for quantifying social impact indicators

Abstract

To make it easier to calculate social impact indicators for evaluating policies. [Solution] The social impact indicator quantification device 103, which quantifies social impact indicators for evaluating policies, includes a logic model construction unit 12 that generates multiple logic models to which weight coefficients are attached, a clustering processing unit 16 that classifies the generated logic models into multiple clusters based on the similarity score indicating the similarity between the generated logic models, a representative logic model selection unit 17 that selects a representative logic model from each of the classified clusters, and a simulation unit 18 that calculates the social impact indicator value of the selected representative logic model as the social impact indicator value of the policy.

Description

Technical Field

[0001] The present invention relates to a technique for quantitatively evaluating social impact indicators.

Background Art

[0002] In recent years, in policy decisions of local governments and investment and financing judgments of financial institutions, etc., it has become important to evaluate how policies and investment and financing affect society. For this evaluation, social impact indicators are used.

[0003] Regarding such evaluations like social impact indicators, a method using a logic model has been proposed.

[0004] For example, in Patent Document 1, regarding communication in sentences and texts, a "communication support system for supporting the development of a logical discussion" is disclosed. Specifically, in Patent Document 1, "a plurality of analysis aspects using a template that satisfies the establishment as a sentence or text are provided, and these are controlled to proceed according to communication basic information", and "a communication support system that has implementation means capable of corresponding to both convergence and divergence in the template and the recording means, and supports a logical discussion development for a topic having a complex causal relationship or a topic having an either-or nature" is disclosed.

Prior Art Documents

Patent Documents

[0005]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0006] Systems like the one described in Patent Document 1 are effective for creating logic models. However, the system in Patent Document 1 is qualitative in its evaluation of social impact indicators, making it difficult to compare the numerous policy candidates that are needed.

[0007] Therefore, the objective of this invention is to more easily calculate social impact indicators for evaluating policies, investments, etc. (simply referred to as policies). [Means for solving the problem]

[0008] To solve the above problems, the present invention classifies multiple logic models related to social impact indicators using their similarity.

[0009] More specifically, the social impact indicator quantification method, which is performed by a social impact indicator quantification device to quantify social impact indicators for evaluating policies, involves a logic model construction unit generating multiple logic models to which weight coefficients are attached, a classification unit classifying the generated logic models into multiple clusters based on a similarity score indicating the similarity between the generated logic models, and a social impact indicator value calculation unit calculating the social impact indicator value of the policy.

[0010] The present invention also includes a social indicator quantification device that performs a method for quantifying social impact indicators, a social impact indicator quantification program that enables the device to function as a computer, and a storage medium that stores the social impact indicator quantification program. [Effects of the Invention]

[0011] According to the present invention, social impact indicators for evaluating policies can be calculated more easily. [Brief explanation of the drawing]

[0012] [Figure 1]This is an explanatory diagram showing an example of the system configuration of a social impact index quantification system 100 according to one embodiment of the present invention. [Figure 2] This figure shows an example of the hardware configuration of each device constituting the social impact index quantification system 100 according to one embodiment of the present invention. [Figure 3] This is a flowchart showing the process for quantifying social impact indicators according to one embodiment of the present invention. [Figure 4] This figure shows a logic model which is a node selected in the node selection process (step S303) according to one embodiment of the present invention. [Figure 5] This figure shows an example of weight coefficients added to a logic model according to one embodiment of the present invention. [Figure 6] This figure shows an example of weight coefficients added to a logic model according to one embodiment of the present invention. [Figure 7] This figure shows the effect on social impact nodes of measures related to one embodiment of the present invention. [Figure 8] This figure shows an example of the impact of a logic model on one embodiment of the present invention. [Figure 9] This figure shows the degree of deviation related to one embodiment of the present invention. [Figure 10] This figure shows a distance matrix relating to one embodiment of the present invention. [Figure 11] This figure schematically shows a clustered logic model according to one embodiment of the present invention. [Figure 12] This diagram shows a representative logic for one embodiment of the present invention. [Figure 13] This figure shows the social impact index values ​​of a representative logic model according to one embodiment of the present invention. [Figure 14] This figure shows an example of displaying the results of a quantitative processing according to one embodiment of the present invention. [Figure 15] This is a functional block diagram of a social impact index quantification device 103 according to one embodiment of the present invention. [Modes for carrying out the invention]

[0013] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. In this embodiment, a case of constructing a logic model aiming at improving social impact indicators, which are long-term goals, in a local government or the like will be described.

[0014] In the social impact indicator quantification system 100 of this embodiment, it supports the formulation of measures to improve the social impact indicators of the target area. For this purpose, in this embodiment, it predicts how much the social impact indicators will change by measures through a logic model. And by using this result, for example, it is possible to provide measures with the highest improvement prediction.

[0015] Therefore, the acceptability of the logic model for users such as local governments becomes an important factor. In this case, for the local government, the logic model of stakeholders including residents is important. Here, the number of logic models may increase in proportion to the number of stakeholders.

[0016] Regarding this point, in the existing social impact indicator quantification system, when dealing with a large number of logic models, the quantitative value of the social impact indicator (hereinafter referred to as the social impact indicator value) is calculated for each logic model, making it difficult for users such as local governments to manage.

[0017] Therefore, in this embodiment, a plurality (for example, a large number) of logic models are reduced in number to representative logic models by clustering using similarity, and this is presented. As a result, it becomes easier for users to make decisions.

[0018] Figure 1 is an explanatory diagram showing an example of the system configuration of the social impact index quantification system 100 according to this embodiment. In Figure 1, the social impact index quantification system 100 consists of a first information terminal 101, a second information terminal 102, a social impact index quantification device 103, and a database 104, all connected to each other via a network 105. The network 105 can be implemented using the Internet, a LAN (Local Area Network), a WAN (Wide Area Network), etc.

[0019] The first information terminal 101 is a computer used by participants 1 to n (where n is an integer greater than or equal to 1; hereinafter collectively referred to as "participants"). The first information terminal 101 receives the structure of the logic model and designated indicators that are deemed important from the participants, and transmits this information to the social impact indicator quantification device 103. The first information terminal 101 can be implemented as a communication device (computer), such as a personal computer or a smart device.

[0020] Furthermore, the second information terminal 102 is an information terminal used by decision-makers such as local governments. The second information terminal 102 is a computer that receives and displays the logic model and simulation results, which are social impact index values, from the social impact index quantification device 103.

[0021] Therefore, the second information terminal 102 can also be implemented as a communication device (computer), such as a personal computer or smart device. Although only one second information terminal 102 is shown in the diagram, multiple terminals may be provided. In this case, an organization such as a local government may provide multiple second information terminals 102, or each organization may have only one.

[0022] Furthermore, the social impact indicator quantification device 103 receives the specified indicators designated by the participant from the first information terminal 101, and constructs a quantifiable logic model from the acquired logic model based on the specified indicators.

[0023] Furthermore, the social impact indicator quantification device 103 performs simulations of quantifiable logic models and information processing related to the quantification of social impact indicators indicated by specified indicators.

[0024] Furthermore, the social impact index quantification device 103 transmits the processing results, such as quantifiable logic models and the results of information processing (e.g., social impact index values), to the second information terminal 102. The database 104 stores the data and processing results used in the processing by the social impact index quantification device 103.

[0025] Next, an example of implementation of each device constituting the social impact index quantification system 100 will be described. Figure 2 is a block diagram showing an example of hardware configuration when each device constituting the social impact index quantification system 100 according to this embodiment is implemented using a computer.

[0026] In Figure 2, the first information terminal 101, the second information terminal 102, and the social impact index quantification device 103, which constitute the social impact index quantification system 100, are collectively referred to as the computer 200.

[0027] Computer 200 includes a processor 201, a storage device 202, an input device 203, an output device 204, and a communication interface (communication IF) 205. The processor 201, storage device 202, input device 203, output device 204, and communication IF 205 are connected by a bus 206.

[0028] Furthermore, the processor 201 controls the computer 200 and performs calculations according to various programs, such as a social impact indicator quantification program. The memory device 202 functions as memory, which serves as the work area for the processor 201. In addition, the memory device 202 is also a non-temporary or temporary recording medium for storing various programs and data.

[0029] As the storage device 202, for example, ROM (Read Only Memory), RAM (Random Access Memory), HDD (Hard Disk Drive), and flash memory can be used.

[0030] Furthermore, the input device 203 accepts input such as data and instructions from the user. Examples of input devices 203 include a keyboard, mouse, touch panel, numeric keypad, scanner, and microphone.

[0031] Furthermore, the output device 204 outputs data such as the calculation results from the processor 201. For example, the output device 204 could be a display, printer, speaker, or the like.

[0032] Furthermore, since the social impact index quantification device 103 can be implemented on a server, the input device 203 and output device 204 can be omitted. In addition, the input device 203 and output device 204 may be integrated into a single unit, such as a touch panel. The communication IF 205 is connected to the network 105 and transmits and receives data with other devices.

[0033] Here, we will describe the functional blocks for realizing the functions of the social impact index quantification device 103 described above. Figure 15 is a functional block diagram of the social impact index quantification device 103 according to this embodiment. As shown in Figure 15, the social impact index quantification device 103 is connected to a network 105 and has a communication unit 11, a logic model construction unit 12, an impact calculation unit 13, a deviation calculation unit 14, a distance calculation unit 15, a clustering processing unit 16, a representative logic model selection unit 17, a simulation unit 18, and an evaluation unit 19.

[0034] Furthermore, in the example shown in Figure 15, the database 104 is configured to be installed in the social impact index quantification device 103.

[0035] First, the communication unit 11 corresponds to the communication IF 205 in Figure 2, connects to the network 105, and sends and receives data with other devices. At this time, it receives information such as the target policies and designated indicators.

[0036] Furthermore, the logic model construction unit 12 constructs the aforementioned logic model. To this end, the logic model construction unit 12 generates multiple logic models to which at least weight coefficients are added.

[0037] Furthermore, the influence calculation unit 13 calculates the degree of influence on the generated logic model. The deviation calculation unit 14 calculates the deviation between the multiple logic models for which influence has been calculated. Here, the deviation indicates the degree of difference or approximation between two logic models, and the details will be explained later.

[0038] Furthermore, the distance calculation unit 15 calculates the distance between the logic models for which the degree of deviation has been calculated. For example, a distance matrix can be used as the distance. In this way, the distance calculation unit 15 calculates a distance corresponding to the degree of influence calculated by the influence calculation unit 13 and the degree of deviation calculated by the degree of deviation calculation unit 14. This distance is an example of the similarity between logic models, and therefore, the distance calculation unit 15 is an example of a similarity calculation unit.

[0039] Furthermore, the clustering processing unit 16 performs clustering on the generated logic models based on the calculated distance. Here, the deviation is an example of an evaluation value that shows the approximation of the formal weights (vector quantities) between logic models. The distance is an example of a similarity value that shows the similarity between logic models.

[0040] Thus, in this embodiment, clustering of the generated logic models is performed based on the degree of divergence, which is an example of the approximation of formal weights (vectors) between logic models, and the degree of influence, which indicates the degree of influence of each node on the final node.

[0041] The similarity metric only needs to indicate the similarity between the logic models. For example, the similarity metric may be based on either approximation or influence. In this case, the similarity metric may be based on at least approximation (more preferably deviation).

[0042] As a result, the clustering processing unit 16 performs clustering of the generated logic models according to at least the degree of similarity corresponding to the degree of deviation.

[0043] Therefore, the influence calculation unit 13, deviation calculation unit 14, distance calculation unit 15, and clustering processing unit 16 of this embodiment are examples of classification units that classify logic models generated based on similarity into multiple clusters. Thus, similarity is an index corresponding to the degree of formal weight approximation between the generated logic models and the degree of influence at each node on the final node.

[0044] Then, the representative logic model selection unit 17 aggregates the logic models according to the clustering results, which is one example of classification. To do this, the representative logic model selection unit 17 aggregates the logic models by selecting a representative logic model from each cluster. In this way, the representative logic model selection unit 17 reduces the number of logic models it references.

[0045] Furthermore, the simulation unit 18 performs a simulation on a representative logic model and quantifies the social impact indicators shown by it. In other words, the simulation unit 18 calculates the social impact indicators for the target policy. Then, the evaluation unit 19 uses the quantified social impact indicator values ​​to evaluate the policy.

[0046] The logic model construction unit 12, impact calculation unit 13, deviation calculation unit 14, distance calculation unit 15, clustering processing unit 16, representative logic model selection unit 17, simulation unit 18, and evaluation unit 19 perform calculations on the processor 201 shown in Figure 2. In other words, the processor 201 executes the processing of the logic model construction unit 12, simulation unit 18, and evaluation unit 19 according to the social impact index quantification program. Details of these processes will be described later.

[0047] Furthermore, the database 104 in Figure 15 stores logic model data 106. The logic model data 106 shows the results of the quantification process and includes the generated logic model, the representative logic model, and its social impact index values. Moreover, it is desirable that the logic model data 106 records the generated logic model and its representative logic model for each cluster.

[0048] Next, the social impact indicators and logic model in this embodiment will be described in detail. While there is generally no clear definition of social impact indicators, in this embodiment they are indicators that show the social impact produced by activities and policies. The social impact indicators cover issues such as equality, living standards, health, nutrition, poverty, safety, and justice. Specific examples in local communities also include subjective indicators such as residents' work ethic and awareness of the SDGs. Furthermore, social impact is increasingly being used in the financial sector, such as ESG investment and lending and social impact bonds, to consider the impact on society and the environment in investment decisions.

[0049] Therefore, it is desirable to evaluate social impact indicators and use the results when formulating policies. There are several methods for evaluating these social impact indicators, one of which is the logic model. The logic model is represented by a network model consisting of nodes and edges, where the nodes represent various indicators and the edges represent the causal relationships between indicators.

[0050] Furthermore, a logic model is a conceptual tool for clarifying the logic of a social indicator quantification program. Specifically, a logic model can clarify how the resources invested will logically influence the target social impact indicator. A typical logic model often consists of five main layers.

[0051] The explanations for each of these layers are provided below.

[0052] Layer 1: This is the input layer, encompassing both resources and constraints for implementing policies. Resources are utilized when policies are implemented and include not only human, material, and financial resources, but also cultural resources. It is crucial to leverage these resources and generate effective social impact indicators within the constraints.

[0053] Layer 2: This is the activity layer, which includes policies or actions and activities for implementing policies, and includes processes, tasks, and actions.

[0054] Layer 3: This is the output layer, representing the direct results of the activity layer carried out by the implementers. These results include business activities for beneficiaries, such as the products and services provided, and ongoing support for the project.

[0055] Layer 4: This is the outcome layer, representing the ripple effects of the output layer described above, and the necessary indicators for improving social impact metrics. "Outcomes" often include subjective indicators such as increased motivation, satisfaction, and awareness as targets of ripple effects. In some cases, short-term and long-term effects are treated separately within "outcomes."

[0056] Layer 5: This is the Impact Layer (Social Impact Layer), which represents the ultimate goal shared by the implementer or all stakeholders, and serves as an improvement indicator toward solving social problems. Broadly speaking, it addresses issues covered by the SDGs, such as reducing inequality, but depending on the organization, more local goals may be set. In addition, output or outcome indicators may be used as substitutes for social impact indicators. In this case, intermediate indicators are considered as targets.

[0057] In this context, conventional logic models primarily focus on logically clarifying ripple effects. Furthermore, these ripple effects, particularly in the outcome layer, include qualitative and subjective indicators. These indicators are necessary from the perspective of the purpose of social impact indicators and appear effective in evaluating the validity of the logic.

[0058] Furthermore, conventionally, logic models have been used in which weight coefficients (numerical values) are added to the edges of the logic model, allowing for the quantification of social impact indicators. These weight coefficients are primarily numerical values ​​that reflect the user's preference order and values.

[0059] Generally, values ​​are considered diverse. Therefore, when there are many users and stakeholders involved in creating logic models, the variations in weight coefficients inevitably increase, resulting in an increase in logic models with the same topology but different weight coefficients. This increase in logic models can make it difficult for the user making the final decision to grasp the overall picture.

[0060] In this embodiment, to address the challenge of the existing quantifiable logic model, which involves an increasing number of logic models with different weight coefficients, a smaller number of representative logic models are selected through logic model clustering. The similarity between the logic models is used in this process. This makes it easier to understand the intentions and reach a consensus, even with multiple, especially numerous, logic models.

[0061] Next, the processing flow of this embodiment will be described. Figure 3 is a flowchart of the social impact index quantification process according to this embodiment. This social impact index quantification process incorporates a logic model clustering function and calculates quantitative values ​​of social impact indicators. The following explanation of each step in Figure 3 will refer to the configuration in Figure 15. Therefore, each step also means that the processor 201 in Figure 2 executes processing according to the social impact index quantification program.

[0062] First, in step S301, the logic model construction unit 12 selects indicators for each of the five layers of the logic model mentioned above. Here, the logic model construction unit 12 treats the quantifiable logic model by grouping the five layers (layers 1 to 5) into three layers that behave differently from a quantifiable perspective. The grouping is done as follows: the input layer is the first layer, the activity layer to the outcome layer is grouped into the second layer (activity / output / outcome), and similarly the impact layer is the third layer (social impact). At this time, the logic model construction unit 12 selects indicators according to the specified indicators received by the communication unit 11, which functions as an input unit.

[0063] For illustrative purposes, in this embodiment, the nodes in each layer are described as follows: the first layer as the driving node, the second layer as the intermediate node, and the third layer as the social impact node. In logic models, each node often represents an indicator or policy.

[0064] For example, if the social impact is the city's landscape, the social impact node might be "good landscape," and related indicators could include "abundance of forests" and "visibility of mountains." Then, proposed measures to achieve a good landscape might include "planting trees." In step S301, the logic model construction unit 12 selects nodes according to these objectives.

[0065] This selection may be carried out in response to instructions received from the first information terminal 101 via the communication unit 11. In this case, it is desirable to select the node with the most participants specified. Furthermore, it is also possible to store previously selected nodes in the database 104 and use that database.

[0066] Next, in step S302, the logic model construction unit 12 creates edges between the selected nodes. To do this, the logic model construction unit 12 connects the nodes selected in step S301 in a tree structure, as shown in the example in Figure 4. This may be done in response to instructions received from the first information terminal 101 via the communication unit 11, similar to step S301.

[0067] Furthermore, edges created in the past can be stored in database 104 and used from there. In this way, edges that represent a graph structure in which nodes are connected in a tree-like manner become a general logic model.

[0068] Next, in step S303, the logic model construction unit 12 adds weights to the edges. At this time, the logic model construction unit 12 adds a weight coefficient to each edge, i.e., each logic model, created in step S302.

[0069] This process, like steps S301 and S302, is performed in response to input instructions from the user via the first information terminal 101. However, in step S303, unlike steps S301 and S302, each user inputs a weight into the first information terminal 101, and the logic model construction unit 12 adds these weights.

[0070] The following is an example. For instance, consider the case where User 1 and User 2 each input weights to each edge of the logic model shown in Figure 4.

[0071] In this case, an example of the weight coefficients added by user 1 is shown in Figure 5. Similarly, the weight coefficients added by user 2 are also shown in Figure 5. First, in Figure 5, Table 501 lists the weight coefficients of the nodes input to intermediate nodes 1-2. Specifically, the weight coefficient assigned to the edge input from drive node 1 is 0.1 according to Table 501, the weight coefficient corresponding to drive node 2 is 0.4, and the weight coefficient corresponding to drive node 3 is 0.5.

[0072] There are various ways to determine weight coefficients, but one example of an existing technology is preference order. A larger weight coefficient indicates that the user considers it more important. In other words, in Table 501, the most important factor influencing intermediate node 1-2 is considered to be drive node 2, followed by drive node 3 and then drive node 1. For example, if intermediate node 1-2 is "forest landscape," then drive node 2 would be "afforestation."

[0073] Furthermore, the weight coefficients for user 2 are shown in Figure 6. A difference is observed when comparing the weight coefficients in Figure 6 with those for user 1 in Figure 5. This represents the difference in the indicators that user 1 and user 2 consider important, i.e., their values. If there were multiple users, similar coefficients could be generated for each user. In other words, after process 303, numerous logic models are generated.

[0074] Here, we will explain the method for quantifying social impact indicators. This quantification method can utilize conventional techniques and is merely one example; other methods may also be used.

[0075] First, the nodes we are looking for are the two nodes shown in the third layer in Figure 4, and we assume that the weight coefficients shown in Figure 5 are assigned to each edge. Furthermore, we assume that the drive node in the first layer is assigned a value between 0 and 1.

[0076] In this example, for the sake of simplicity, we will provide three numerical examples. The combination of numerical values ​​given to the drive nodes will be referred to as a "measure." Measure 1 consists of three examples: drive node 1=1, drive node 2=0, drive node 2=0; Measure 2 consists of drive node 1=0, drive node 2=1, drive node 2=0; and Measure 3 consists of three examples: drive node 1=0, drive node 2=0, drive node 2=1.

[0077] After inputting a numerical value into the drive node, the numerical value of the intermediate node is calculated. As an example, consider intermediate node 1-2. The weight coefficients are as shown in Table 501 in Figure 5. The value of intermediate node 1-2 is given by the following (Equation 1). The value of intermediate node 1-2 = the sum of (each weight coefficient × the numerical value of each node input to intermediate node 1-2) ... (Equation 1) Therefore, in policy 1, the value of intermediate node 1-2 is calculated as = 1 × 0.1 + 0 × 0.4 + 0 × 0.5.

[0078] This process is then repeated for the other nodes to calculate the values ​​for the third layer nodes. Figure 7 shows the values ​​obtained by normalizing these numbers (the smallest number is 0, and the largest number is 1). This allows us to obtain numerical values ​​representing the effect on the social impact nodes based on how the policies are weighted.

[0079] This concludes the explanation of weighting in step S303. Returning to Figure 3, we will continue the explanation of the social impact index quantification process. Next, we will perform a reduction method to reduce the number of logic models to which weight coefficients have been added, as generated as described above. This reduction method is performed in the calculation of impact (step S304), the calculation of deviation (step S305), the calculation of distance (step S306), the clustering process (step S307), and the calculation of representative logic models (step S308).

[0080] The following explains each step. For the purposes of this explanation, we will assume that weight coefficients have been added up to step S303, meaning that 10 logic models have been generated. We will also refer to User 1's logic model as Logic Model 1, User 2's as Logic Model 2, and so on.

[0081] First, in step S304, the impact calculation unit 13 quantifies the impact of the second-layer nodes that have a large degree of influence on the third-layer social impact nodes in each generated logic model. A large degree of influence means that it is greater than a predetermined standard impact, greater than other nodes, or otherwise meets predetermined criteria.

[0082] Figure 8 shows examples of the impact of Logic Model 1 and Logic Model 2. The impact represents the change in the numerical value of the social impact node when the value of a specific intermediate node is fixed to 0, compared to the calculation method for quantifying the social impact index shown earlier.

[0083] For example, the value of the social impact node when all inputs to the drive node are set to 0.5 is used as the baseline. Then, the impact calculation unit 13 calculates the change in the value of the social impact node when the value of intermediate node 2-1 is fixed to 0 as the impact of intermediate node 2-1. The impact calculation unit 13 performs the same calculation for the other intermediate nodes in the second layer and can obtain a graph like the one in Figure 8.

[0084] In this way, the influence calculation unit 13 calculates the influence of each node in the target logic model. Alternatively, the unit may be configured to calculate the influence of intermediate nodes in the target logic model.

[0085] Next, in step S305, the deviation calculation unit 14 calculates the deviation between the generated logic models. This involves calculating the deviation between each generated logic model and other logic models. This deviation quantifies the difference between two logic models and is an example of approximation.

[0086] This embodiment will be described as an example of calculating the degree of discrepancy between logic model 1 and logic model 2. First, the degree of discrepancy is calculated for each node of the intermediate node in the second layer and the social impact node in the third layer. The following will be a specific explanation using intermediate node 1-2 as an example.

[0087] First, the weight coefficients of each node input to intermediate node 1-2 of logic model 1 are 0.1, 0.4, and 0.5. Similarly, the weight coefficients of each node input to intermediate node 1-2 of logic model 2 are 0.3, 0.4, and 0.3.

[0088] In this case, the nodes input to intermediate node 1-2 of logic model 1 and intermediate node 1-2 of logic model 2 will have the same number. In this case, these are drive node 1, drive node 2, and drive node 3.

[0089] In this embodiment, the input node is the node immediately preceding the target intermediate node, and in addition to the driving node, the (previous) intermediate node may also be input.

[0090] Therefore, the deviation calculation unit 14 calculates the cosine similarity using the set of weight coefficients as a vector. That is, the set of weight coefficients is {0.1, 0.4, 0.5} for vector 1 and {0.3, 0.4, 0.5} for vector 2.

[0091] The cosine similarity in this case is 0.90. Furthermore, since what we want to find is the degree of deviation, the degree of deviation calculation unit 14 divides by 1, as it has the inverse meaning of similarity, and sets the degree of deviation for that node to 1-0.9=0.1. The degree of deviation calculation unit 14 similarly calculates the degree of deviation for other intermediate nodes. An example of the degree of deviation calculated in this way is shown in Figure 9. In this way, the degree of deviation calculation unit 14 calculates the degree of deviation for each node of the target logic model. Note that the configuration may also be used to calculate the degree of deviation for intermediate nodes of the target logic model.

[0092] Furthermore, cosine similarity can also be calculated using other formulas that calculate the distance between vectors, such as the sum of absolute differences or the sum of squared differences.

[0093] Next, in step S306, the distance calculation unit 15 calculates the distance between the generated logic models. This involves calculating the distance between each generated logic model and other logic models. This distance is calculated between two logic models and is derived from the degree of influence and the degree of deviation.

[0094] The following explains how to calculate the distance between node i in logic model 1 and logic model 2. First, the distance calculation unit 15 calculates the distance to each intermediate node as (influence of node i in logic model 1 + influence of node i in logic model 2) / 2 × (deviation of node i in logic model 1 and logic model 2).

[0095] Then, the distance calculation unit 15 calculates the sum of the calculated distances for each intermediate node as the distance between logic model 1 and logic model 2.

[0096] As described above, the distance in this embodiment is calculated from the degree of influence and the degree of deviation, but it may also be calculated from the degree of influence alone. In this case, for example, the distance is calculated as (degree of influence of node i in logic model 1 + degree of influence of node i in logic model 2) / 2. This concludes the explanation of how to calculate the distance between logic model 1 and logic model 2.

[0097] In this embodiment, we assume there are 10 logic models. Of these, the details of distance calculation in logic models 3 to 10 can be implemented in the same way as logic models 1 and 2, with only the weight coefficients differing, so we will omit the explanation.

[0098] As described above, when the distance is calculated for 10 logic models, the resulting distance is represented by the distance matrix shown in Figure 10. In Figure 10, the logic model number of each logic model is plotted on the vertical and horizontal axes, and the intersection of these axes represents the distance between the two logic models with the corresponding logic model numbers.

[0099] Thus, the distance between logic model 1 and logic model 2 is the value at the intersection of their respective rows and columns. Therefore, for 10 logic models, the distance matrix will be 10 rows and 10 columns. In other words, if n logic models are generated, the distance matrix will be n rows and n columns. This concludes the explanation of distance calculation in step S306.

[0100] Next, we will explain the clustering in step S307. In step S307, the clustering processing unit 16 performs clustering based on the distance calculated in step S306, i.e., the distance matrix. As a result, the generated logic model is classified into multiple clusters. In this embodiment, a dendrogram algorithm as shown in Figure 11 was adopted as the clustering algorithm.

[0101] Figure 11 schematically shows the clustered logic model, which is the result of processing up to step S306. In Figure 11, the horizontal axis represents the logic model number of the clustered logic model, and the vertical axis represents the distance between logic models (subclusters) connected by line segments, i.e., the combination distance in the subclusters. This distance will be explained later.

[0102] Then, the clustering processing unit 16 clusters logic models that are close to each other in a tree-like structure. To do this, the clustering processing unit 16 first creates subclusters, which are pairs of adjacent logic models. Here, "adjacent" means the closest distance between them. In the example in Figure 11, subclusters are created for the pairs of logic models 3 and 8, 5 and 7, 1 and 10, 2 and 9, and 4 and 6.

[0103] Furthermore, the clustering processing unit 16 identifies adjacent subclusters for each subcluster of the created logic model. For example, in Figure 11, subclusters with cluster numbers 2 and 9 are identified as adjacent subclusters to subclusters with logic model numbers 1 and 10.

[0104] Then, the clustering processing unit 16 combines the subclusters to form a new subcluster if the constraints are met. The following constraints can be used: • Is the combination distance in the subcluster within a predetermined value? • Is the distance between subclusters within a predetermined distance? • Is the number of subcluster combinations greater than or equal to a predetermined number? • Is the number of layers with the highest number in a subcluster greater than or equal to a predetermined value? • Is the number of subclusters created less than or equal to a predetermined number? Furthermore, at least some of these may be combined and used as constraints.

[0105] In the example shown in Figure 11, the combination distance is used as a constraint. This combination distance represents the distance obtained when combining logic models or subclusters, including the distances up to that point.

[0106] The clustering processing unit 16 then calculates the distance between each candidate subcluster in order to identify adjacent subclusters. This is calculated in the same way as the distance between logic models. In the example shown in Figure 11, the distance between the virtual intermediate nodes of each candidate cluster is calculated by (influence of node i of logic model 3 + influence of node i of logic model 8 + influence of node i of logic model 5 + influence of node i of logic model 7) / 4 × (deviation of node i of logic model 3, logic model 8, logic model 5 and logic model 7). The sum of these distances is then used to calculate the distance between the candidate clusters.

[0107] These calculations may be performed by the clustering processing unit 16, or by the distance calculation unit 15 upon request from the clustering processing unit 16. The clustering processing unit 16 then identifies adjacent subclusters based on the calculated distances.

[0108] Then, if the distance satisfies the constraints, the clustering processing unit 16 combines these candidate subclusters to form a new subcluster.

[0109] In the example shown in Figure 11, as described above, the subclusters of logic model numbers 1 and 10 and the subclusters of logic model numbers 2 and 9 are combined to form a new subcluster. The clustering processing unit 16 then performs the same processing on the subclusters until it is no longer possible to create a subcluster that satisfies the constraints. In Figure 11, the constraint is that the distance must be within 4.

[0110] The clustering processing unit 16 identifies each subcluster that no longer satisfies the constraints as a cluster. In other words, it clusters the generated logic model.

[0111] Therefore, for example, in the example shown in Figure 11, since the constraint is that the distance is within 4, the 10 generated logic models are clustered into cluster 1 and cluster 2. Note that the constraint may be uniform or variable depending on the hierarchy (number of times) of the subcluster combinations. This concludes the explanation of clustering in step S307.

[0112] Next, we will explain the identification of the representative logic model in step S308. In this step, the representative logic model selection unit 17 selects the representative logic model for each cluster from the clustering results in step S307.

[0113] In this embodiment, the representative logic model selection unit 17 selects a representative logic model using the average value of each cluster. That is, the logic model closest to the average value is selected.

[0114] Here, the mean is just one example of a representative value for a cluster; the median or mode may also be used, and the logic model closest to these values ​​may be selected. Furthermore, at least one of the various calculated values ​​mentioned above, such as influence, deviation, weight, distance or similarity to other logic models, can be used as a representative value.

[0115] As a result, for the logic models clustered into cluster 1 and cluster 2 shown in Figure 11, the representative logic model for cluster 1 is logic model 10, and the representative logic model for cluster 2 is logic model 6, as shown in Figure 12.

[0116] As a result of the above process, the 10 logic models will be consolidated into 2 logic models. In other words, the number of logic models to be targeted will be reduced. In this example, we consolidated into 2 clusters using clustering, but this number can be changed arbitrarily. This concludes the explanation of identifying the representative logic model in step S308.

[0117] Next, in step S309, the simulation unit 18 calculates the social impact index value for the selected representative logic model. As a result, the social impact index of the logic model is quantified. In this embodiment, the social impact index value for the social impact node of the representative logic model is calculated as the social impact index value for the target policy.

[0118] In this embodiment, the simulation unit 18 calculates the social impact index value by executing a simulation, but the social impact index value may be calculated by other methods. For this reason, the simulation unit 18 is just one example of a social impact index value calculation unit.

[0119] Furthermore, the social impact index calculation unit only needs to calculate the social impact index value for the target policy based on the clustering results from the clustering processing unit 16, for example, based on the clusters, and does not necessarily need to use a representative logic model.

[0120] Figure 13 shows the social impact index values ​​for logic model 10 and logic model 6, which are representative logic models resulting from this clustering.

[0121] From these results, it can be understood that for the logic model belonging to cluster 1, policy 1 is the most effective in terms of social impact, and for the logic model belonging to cluster 2, policy 2 is the most effective. Here, an example of displaying the results of the quantification process according to this embodiment is shown in Figure 14.

[0122] In this embodiment, the display screen 141 in Figure 14 is displayed on a display device, which is an example of an output device 204 of the second information terminal 102. However, it may also be displayed on the display device of the first information terminal 101 or the social impact index quantification device 103.

[0123] In Figure 14, the display screen 141 includes a clustering result display unit 142 and a social impact display unit 143. First, the clustering result display unit 142 displays a tree diagram showing the clustered logic model described in Figures 11 and 12. This clustered logic model may also be displayed in a diagram other than a tree diagram, or the logic model number and cluster number may be summarized in a table format.

[0124] Furthermore, the social impact display unit 143 displays the social impact index values ​​for each cluster, that is, the social impact index values ​​of the representative logic model. The social impact display unit 143 may also display the social impact index values ​​of the representative logic model in a chart, as shown in Figure 13. In addition, the cluster number and the representative logic model for each cluster can also be displayed on this screen.

[0125] The display screen 141 above allows decision-makers to understand, by viewing the results of the quantification process, which measures are suitable for which groups, and how the appropriate measures differ depending on the group. This serves as an aid in decision-making.

[0126] This concludes the description of this embodiment. According to this embodiment, appropriate social impact indicators can be calculated more easily, and differences in policies considered important due to differences in values ​​can be grasped. In particular, the number of logic models to be referenced can be reduced, making it easier for decision-makers to evaluate. Thus, this embodiment makes it possible to evaluate policies more appropriately and to support consensus building regarding policies, including policy decisions. Note that this embodiment is not limited to the example described above, and quantification processing may also be performed on the first information terminal 101 or the second information terminal 102. [Explanation of Symbols]

[0127] 11...Communication Unit, 12...Logic Model Construction Unit, 13...Impact Calculation Unit, 14...Degree of Deviation Calculation Unit, 15...Distance Calculation Unit, 16...Clustering Processing Unit, 17...Representative Logic Model Selection Unit, 18...Simulation Unit, 19...Evaluation Unit, 100...Social Impact Indicator Quantification System, 101...First Information Terminal, 102...Second Information Terminal, 103...Social Impact Indicator Quantification Device, 104...Database, 105...Network, 106...Logic Model Data, 200...Computer, 201...Processor, 202...Storage Device, 203...Input Device, 204...Output Device, 205...Communication Interface

Claims

1. In a method for quantifying social impact indicators, which is performed by a social impact indicator quantification device, for quantifying social impact indicators used to evaluate policies, The logic model construction unit generates multiple logic models to which weight coefficients are added. The classification unit classifies the generated logic models into multiple clusters based on the similarity score indicating the similarity between the generated logic models. A method for quantifying social impact indicators, wherein a social impact indicator value calculation unit calculates the social impact indicator value of the aforementioned policy.

2. In the method for quantifying social impact indicators according to claim 1, The classification unit includes a deviation degree calculation unit and a clustering processing unit. The deviation calculation unit calculates a deviation that indicates the degree of approximation of the formal weights between the generated logic models, A method for quantifying social impact indicators, wherein the clustering processing unit classifies the generated logic model into a plurality of clusters based on the similarity corresponding to the degree of deviation.

3. In the method for quantifying social impact indicators according to claim 2, The classification unit has an impact calculation unit, The influence calculation unit calculates an influence level that indicates the degree of influence on the final node of the generated logic model, A method for quantifying social impact indicators, wherein the clustering processing unit classifies the generated logic model into a plurality of clusters based on the degree of divergence and the degree of influence.

4. In the method for quantifying social impact indicators according to claim 1, The classification unit has a distance calculation unit, A method for quantifying a social impact index, wherein the distance calculation unit calculates the distance of the generated logic model as the similarity, corresponding to the sum of the distances of each intermediate node.

5. In the method for quantifying social impact indicators according to claim 1, The aforementioned social impact indicator quantification device includes a representative logic model selection unit, The representative logic model selection unit selects a representative logic model from each of the classified clusters, A method for quantifying social impact indicators, wherein a social impact indicator value calculation unit calculates the social impact indicator value in the selected representative logic model as the social impact indicator value of the measure.

6. In a social impact indicator quantification device that quantifies social impact indicators for evaluating policies, A logic model construction unit that generates multiple logic models to which weight coefficients are added, A classification unit that classifies the generated logic models into multiple clusters based on a similarity score indicating the similarity between the generated logic models, A social impact index quantification device having a social impact index calculation unit that calculates a social impact index value for the aforementioned measures.

7. In the social impact index quantification device according to claim 6, The classification unit includes a deviation degree calculation unit and a clustering processing unit. The deviation calculation unit calculates a deviation that indicates the degree of approximation of the formal weights between the generated logic models, A social impact index quantification device in which the clustering processing unit classifies the generated logic model into the plurality of clusters based on the similarity corresponding to the degree of deviation.

8. In the social impact index quantification device according to claim 7, The classification unit has an impact calculation unit, The influence calculation unit calculates an influence level that indicates the degree of influence on the final node of the generated logic model, A social impact index quantification device in which the clustering processing unit classifies the generated logic model into the plurality of clusters based on the degree of divergence and the degree of influence.

9. In the social impact index quantification device according to claim 6, The classification unit has a distance calculation unit, A social impact index quantification device in which the distance calculation unit calculates the distance of the generated logic model according to the sum of the distances of each intermediate node as the similarity.

10. In the social impact index quantification device according to claim 6, Furthermore, it includes a representative logic model selection unit that selects a representative logic model from each of the classified clusters, A social impact index quantification device in which a social impact index value calculation unit calculates the social impact index value in the selected representative logic model as the social impact index value of the measure.

11. A social impact indicator quantification device is a computer that quantifies social impact indicators for evaluating policies. A logic model construction unit that generates multiple logic models to which weight coefficients are added, A classification unit that classifies the generated logic models into multiple clusters based on a similarity score indicating the similarity between the generated logic models, A social impact index quantification program that functions as a social impact index calculation unit for calculating the social impact index value of the aforementioned measures.

12. In the social impact index quantification program described in claim 11, The classification unit includes a deviation degree calculation unit and a clustering processing unit. The deviation calculation unit calculates a deviation that indicates the degree of approximation of the formal weights between the generated logic models, A social impact index quantification program in which the clustering processing unit classifies the generated logic model into the plurality of clusters based on the similarity corresponding to the degree of deviation.

13. In the social impact index quantification program described in claim 12, The classification unit has an impact calculation unit, The influence calculation unit calculates an influence level that indicates the degree of influence on the final node of the generated logic model, A social impact index quantification program in which the clustering processing unit classifies the generated logic model into the plurality of clusters based on the degree of divergence and the degree of influence.

14. In the social impact index quantification program described in claim 11, The classification unit has a distance calculation unit, A social impact index quantification program in which the distance calculation unit calculates the distance of the generated logic model according to the sum of the distances of each intermediate node as the similarity.

15. In the social impact index quantification program described in claim 11, Furthermore, it functions as a representative logic model selection unit that selects a representative logic model from each of the classified clusters. A social impact index quantification program in which a social impact index value calculation unit calculates the social impact index value in the selected representative logic model as the social impact index value of the measure.

Images