Search

The use of artificial intelligence (AI) in manufacturing has become increasingly common, but the lack of transparency and interpretability of AI models can limit their adoption in critical applications, due to lack of human understanding. Explainable AI (XAI) has emerged as a solution to this problem by providing insights into the decision-making process of AI models. This paper analyses different approaches for AI transparency and interpretability, starting from explainability by-design to post-hoc explainability, pointing out trade-offs, advantages and disadvantages of each one, as well as providing an overview of different applications in manufacturing processes. The paper concludes presenting XMANAI as an innovative platform for manufacturing users to develop insightful XAI pipelines that can assist in their everyday operations and decision-making. The comprehensive overview of methods serves as the ground basis for the platform draft catalogue of XAI models.

The development of Artificial Intelligence (AI) systems in Industry 4.0 has gained momentum due to their potential for increasing efficiency and productivity. However, AI systems can be just as complex and opaque, leading to concerns about their reliability, trustworthiness, and accountability. To address these issues, this paper proposes a validation framework for Explainable AI (XAI) in Industry 4.0 based on Fuzzy Cognitive Maps (FCMs). The proposed framework aims to evaluate Key Performance Indicators (KPIs) based on a set of AI metrics and XAI metrics. The FCM-based approach enables the representation of causal-effect relationships between the different concepts of the system using expert knowledge. The presented validation framework provides a theoretical background for evaluating and optimizing the business values of AI systems based on multiple criteria in the manufacturing industry, demonstrating its effectiveness. The main contributions of this paper are: i) the development of an FCM-based validation framework for XAI in Industry 4.0; ii) the identification of relevant AI and XAI metrics for the evaluation of the KPIs of the theoretical graph model; and iii) the demonstration of the effectiveness of the proposed framework through a case study. The results of this study provide valuable insights into the importance of considering not only accuracy but also efficiency and transparency when developing AI pipelines that generate higher business value. Overall, this paper offers a theoretical foundation and practical insights for organizations seeking to evaluate the business values of their AI systems in Industry 4.0. It emphasizes the importance of explainability and the integration of AI and XAI metrics in achieving transparent and accountable AI solutions that deliver optimal results for the manufacturing industry and beyond.

Fuzzy cognitive maps (FCMs) have gained popularity within the scientific community due to their capabilities in modelling and decision making for complex problems. However, despite the large number of papers presenting advances in mathematical formulation and applications of FCMs, along with some recent tools for this soft computing technique, there is a lack of open source tools with sufficient flexibility for modelling and inference in diverse application domains. Filling this gap, this paper presents an open source package in R programming language, called the ‘fcm’ package, which is able to do scenario analyses and to examine and estimate the inference of FCM, using the fcm.infer function. Six different inference rules (kosko, modified-kosko, rescale rule, and the clamped versions of these rules) and four threshold functions (bivalent, trivalent, sigmoid and hyperbolic tangent) are provided. This open-source package is available in CRAN and is relatively easy to use, even for less experienced users. It thus provides the opportunity for researchers in different fields to analyse their weighted matrices as FCM. This paper shows examples and visualizations to demonstrate the proposed open source FCM package for environmental modelling and decision making.

Carbon Capture and Utilization (CCU) technologies capture CO2 waste emissions and utilize them to generate new products (e.g., fuels, chemicals and materials) with various environmental, economic and social opportunities. However, as most of these CCU technologies are in the R&D stage, their technical and economic viability are examined with very little attention to the social aspects. Besides, the lack of systematic research into social impacts is mainly due to the difficulty in identifying as well as quantifying social aspects through the entire life cycle of CCU products. The first step within social life cycle assessment (SLCA) is to identify the relevant social indicators. Within SLCA it is common to involve stakeholders in indicating which social impact categories and related indicators are most relevant to them. As there are multiple social indicators and stakeholders’ opinions, the identification step is a multi-criteria decision making issue. To address this, in this study a Multi-Criteria Decision Making Technique called TOPSIS (Technique for Order of Preference by Similarity to Ideal Solution) is applied which is rarely utilized for the assessment of social performances within the biobased economy. TOPSIS is employed to empirically determine the relative importance of the indicators for measuring social impacts. To do so, first the relevant stakeholders, potential social impact categories and subcategories, and potential performance indicators are listed using UNEP/SETAC guidelines. Second, through an online questionnaire survey, CCU experts at international level provide linguistic instead of numerical ratings to the social sub-categories and their potential indicators. Afterwards, TOPSIS is used to generate aggregate scores for impact subcategories and to identify indicators of high importance. Finally, sensitivity analysis is performed to determine the influence of criteria weights on the decision making process. The final set of main social indicators resulting from our study provides the basis forthe next steps in the social sustainability assessment of CCU technologies, i.e. data collection and impact assessment. Furthermore, our outcomes can be also used to inform the producers regarding the most and least important social issues for CCU technologies so that the potential social impacts caused by their production activities can be improved or prevented.

Today, more and more businesses are growing constantly trying to stand and maintain in the forefront of their competitive advantage by devising various ways to satisfy their customers or even to create a better "image" to make the company more attractive not only to the consumers but also to the future employees. The difficulty in this case is not the collection of the participants' preferences, but how the conducted knowledge will be reflect better the participants' consensus. Therefore, based on the above, each company will be more competitive, if the factors that influenced the company will be improved. In this paper, Factor Analysis was performed to ascertain which of the given factors are the most important. Then, Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) was used to rank and find the most significant factors that influence the public opinion when they are searching for a job, without expert knowledge. Two other Multi-criteria Decision Making methods the Weighted Sum Model (WSM) and Sum Ranking System (SRS), were considered to rank the factors and then all ranking results were compared. Finally, an aggregation of all evaluations was accomplished, extracting the people consensus. Keywordsrankings; aggregation; MCDA; WSM; TOPSIS.

Carbon capture and utilization (CCU) technologies capture CO2 waste emissions and utilize them to generate new products (such as fuels, chemicals, and materials) with various environmental, economic, and social opportunities. As most of these CCU technologies are in the R&D stage, their technical and economic viability are examined with less attention to the social aspect which is an important pillar for a holistic sustainability assessment. The lack of systematic social impact research is mainly due to the difficulty of identifying and quantifying social aspects through the entire life cycle of products. We will fill this gap for CCU technologies and identify the main social indicators. A multi-criteria decision making tool: TOPSIS (technique for order of preference by similarity to ideal solution) was applied to empirically determine which indicators are more relevant for assessing the social impact of a company operating CCU activities within a European context. First, seeing that social impact categories are linked to key stakeholder groups, we considered workers, consumers, and local communities as relevant stakeholders. Second, the main social impact categories and their potential performance indicators associated to each group of stakeholders were listed using the United Nations Environment Program/Society of Environmental Toxicology and Chemistry (UNEP/SETAC) guidelines. In the third step, an online questionnaire was distributed to identify the main social categories and indicators for CCU, to which 33 European CCU experts responded. Finally, a modified TOPSIS was applied to rank the indicators based on their relevance. We found that the indicators related to “end of life responsibility” and “transparency” within a CCU company achieved the highest rank affecting the consumers group, whereas “fair salary” and “equal opportunities/discriminations” were determined as the most relevant impact categories for the workers. For the local community group, “secure living conditions” and “local employment” received the highest priority from the experts’ point of view. Furthermore, “health and safety” considerations were identified as one of the most important criteria affecting all three groups of stakeholders. The ranking list of the main social indicators identified in our study provides the basis for the next steps in the social sustainability assessment of CCU technologies; that is, data collection and impact assessment. Our outcomes can also be used to inform the producers regarding the most and least relevant social aspects of CCU so that the potential social impacts caused by their production activities can be improved or prevented. We would also like to thank James Morrison from the JamesEdits agency for proof reading and editing the article

Many researchers have studied about attractiveness of a company in business domain but survey after survey, reveals that there is always room for improvement. In this paper, human reasoning extraction is investigated to evaluate the attractiveness of each company resident in Belgium. The term attractiveness refers to the case that a company X pays attention only to customers’ preference to boosting employee satisfaction, help the company retain personnel and attract new employees. This leads to brand name improvement which allows the company to increase the sales of their products or services, remain competitive in the market and increase the employee productivity. These effects can be achieved if the company will focus and improve the factors that participants considered more important. Three Multi-Criteria Decision Making methods: Technique for Order Preference by Similarity to Ideal Solution (TOPSIS), Analytical Hierarchy Process (AHP) and Weighted Sum Model (WSM) were deployed to identify and order the most important factors that influence the company competitiveness based in customer satisfaction. The challenging part of this study is the exploitation of "pure" knowledge from participants, the comparison of results and finally, the aggregation of all accomplished evaluations, without expert knowledge and consequently without weights and criteria. To accomplish this goal, we customized TOPSIS and AHP methods to deal with participants' consensus, without using common voting methods but methods based on MCDM. 14.585 questionnaires were gathered from people in Belgium and 349 companies, which were resident in Belgium, participated in this research. It is important to note that the respondents didn’t have any information about the name of each company. The most significant factors were selected from respondents on the assumption that they wished to choose a company X to be employed in. The questionnaire was divided into two parts, Data Set 1 (DS1) and Data Set 2 (DS2). In DS1, participants gave their preference value only to five of the seventeen factors that consider more important. In DS2, participants had not any limitation for the factors' choice. Each participant had to split the amount of one thousand points to factors that they considered most important by giving more points to the most significant factor. Before ordering the factors, the first procedure was to clean the data in order to achieve the best results. As mentioned before, there is no knowledge about criteria, which are important to determine the ranking, especially in TOPSIS and AHP method. In this paper, each horizontal row of the decision matrix is allocated to one factor and each vertical column to one participant’s opinion. All participants had equal importance so no weights were required. The results have shown that rankings of seventeen factors (of thousands of participants’ opinions which were distributed in fifteen sectors), were similar in three methods. The five top factors that each company is interested to improve were at the top of the list. Since TOPSIS and AHP method was proved effective in our problem to rank properly the factors, we applied once again TOPSIS and AHP method to aggregate the people consensus in the final classification. In these cases, seventeen factors were used as an input (rows) in the decision matrix and fifteen sectors were inserted as columns and applied in DS1 and DS2. Fifteen rankings (one for each sector), that have been ordered with TOPSIS method, were aggregated into a final order, using TOPSIS method and other fifteen rankings that have been ordered with TOPSIS method, were aggregated into a single one, using AHP method. As noted in the final standings the top five factors or the most significant factors, are common in both different data sets. The purpose of the final ranking was to aggregate the common opinion (or else the consensus) and reflect the significant factors that need to improve when the company X cares about the attractiveness or wants to enhance the competitive advantage. The ranking of top 5 factors show that people prefer long term job security and competitive salary package more than offering of interesting jobs, financially sound and pleasant working environment. It also showed that people in Belgium prefer stability (long-term security) instead of jobs financially sound. However, this outcome is not surprising and it could be a direct result of the economic crisis.

Aggregating the preference of multiple experts is a very old problem which remains without an absolute solution. This assertion is supported by the Arrow's theorem: there is no aggregation method that simultaneously satisfies three fairness criteria (non-dictatorship, independence of irrelevant alternatives and Pareto efficiency). However, it is possible to find a solution having minimal distance to the consensus, although it involves a NP-hard problem even for only a few experts. This paper presents a model based on Ant Colony Optimization for facing this problem when input data are incomplete. It means that our model should build a complete ordering from partial rankings. Besides, we introduce a measure to determine the distance between items. It provides a more complete picture of the aggregated solution. In order to illustrate our contributions we use a real problem concerning Employer Branding issues in Belgium.