A Systematic Review of Sustainable Supplier Selection Using Advanced

Artiﬁcial Intelligence Methods

Hanen Neji

, Mouna Rekik

, Lotﬁ Souiﬁ

and Ismail Bouassida Rodriguez

ReDCAD, ENIS, University of Sfax, Tunisia

MIRACL, ISIMS, University of Sfax, Tunisia

ﬁ ﬁ

Keywords:

Sustainable Supplier Selection, Sentiment Analysis, Text Analytics, Multi-Criteria Decision-Making

(MCDM).

Abstract:

Artiﬁcial intelligence (AI) algorithms have signiﬁcantly advanced various ﬁelds, driving innovation in domains

such as healthcare, ﬁnance, and sustainability. In the realm of sustainable development, selecting suppliers

is crucial for promoting environmental responsibility and safeguarding the well-being of future generations.

This complex decision-making process requires evaluating suppliers across numerous criteria. Multi-Criteria

Decision-Making (MCDM) and AI techniques, including Natural Language Processing (NLP), Deep Learn-

ing (DL), and Machine Learning (ML), have emerged as powerful tools to address these challenges. However,

these methods often face transparency issues and the risk of greenwashing, which can erode trust in sustain-

ability assessments. To address this, we conducted a systematic literature review (SLR) of 44 papers published

between 2019 and 2024, sourced from databases such as Springer (12 papers), IEEE Xplore Digital Library

(11 papers), and Science Direct (21 papers). This review offers an equitable analysis of MCDM and AI mod-

els (NLP, DL, ML) for evaluating both supplier sustainability and the risk of greenwashing. Additionally,

sentiment analysis techniques are integrated to enhance transparency and provide insights into stakeholder

perceptions.

1 INTRODUCTION

In the rapidly evolving landscape of sustainability,

the ability to effectively analyze and interpret vast

amounts of data has become crucial. With the increas-

ing complexity of sustainability assessments, leverag-

ing advanced technologies and methodologies is es-

sential for making informed decisions. Cutting-edge

approaches in data analysis and artiﬁcial intelligence

are pivotal in enhancing our understanding of sustain-

ability practices, particularly in evaluating supplier

performance across various dimensions.

In the current global context, heightened environ-

mental concerns have propelled the selection of sus-

tainable suppliers to the forefront of corporate pri-

orities. Yet, evaluating these suppliers solely based

on sustainable criteria reveals inherent imperfections,

emphasizing the pressing need for a comprehen-

sive assessment of their sustainability. This imper-

ative stems from the timeless deﬁnition outlined by

the United Nations Brundtland Commission in 1978,

which advocates for meeting present needs without

compromising the ability of future generations to

meet their own.

Currently, evaluating supplier sustainability based

on (Environmental, Social, Governance) ESG crite-

ria and the Triple Bottom Line (TBL)—encompassing

economic, social, and environmental dimensions—is

gaining momentum as the gold standard. This ap-

proach offers a comprehensive perspective on sup-

plier performance, including environmental steward-

ship, social responsibility, and transparent governance

practices.

However, the multifaceted nature of these crite-

ria presents a complex decision-making landscape,

prompting the adoption of Multi-Criteria Decision-

Making (MCDM) methods to facilitate the selection

of sustainable suppliers. MCDM methods system-

atically evaluate multiple, often conﬂicting, criteria

such as environmental impact, social responsibility,

and economic performance. By using quantitative

and qualitative data, MCDM helps prioritize suppliers

who best align with sustainability goals, minimizing

subjective biases through a structured framework.

Machine Learning (ML) and Deep Learning (DL)

techniques further enhance the objectivity of this se-

Neji, H., Rekik, M., Souiﬁ, L. and Rodriguez, I. B.

A Systematic Review of Sustainable Supplier Selection Using Advanced Artiﬁcial Intelligence Methods.

DOI: 10.5220/0013149900003890

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 17th International Conference on Agents and Artiﬁcial Intelligence (ICAART 2025) - Volume 3, pages 451-460

ISBN: 978-989-758-737-5; ISSN: 2184-433X

451

lection process by analyzing vast amounts of data

from social media and sustainability reports with high

precision and consistency. These technologies iden-

tify patterns and insights that might be missed by

human evaluators, enhancing the objectivity and ef-

ﬁciency of ESG criteria analysis in supplier evalua-

tions.

Nevertheless, reliance solely on objective evalua-

tion methods poses risks, such as greenwashing and

opacity. Greenwashing refers to the practice where

companies misleadingly present their sustainability

initiatives to maintain a positive image while maxi-

mizing proﬁts, often at the expense of transparency

and genuine environmental impact (Vinella et al.,

2023).

Herein lies the signiﬁcance of sentiment anal-

ysis as a complementary method, augmenting the

credibility and transparency of supplier evaluations.

By providing insights into stakeholders’ perceptions

and sentiments, sentiment analysis enriches evalua-

tion processes, fostering a more balanced and trans-

parent approach to supplier sustainability assessment.

Moving forward, in Section 2, we outline the plan-

ning of our Systematic Literature Review, followed by

the presentation of Results and discussion in Section

3. Finally, we conclude with overarching highlights

and propose avenues for further research in Section 4.

2 SYSTEMATIC LITERATURE

REVIEW PLANNING

2.1 Research Questions

Aiming to explore the application of web-based and

data-driven methodologies in sustainable supplier se-

lection and greenwashing detection, the following re-

search questions (RQ) were established:

• RQ1. What criteria and web-based data sources

are essential for evaluating sustainable supplier?

• RQ2. What advanced methods and data science

approaches for sustainable supplier selection?

• RQ3. How can web-driven sentiment analysis and

text analytics contribute to detecting greenwash-

ing?

Subsequently, we determined the initial research in

the research databases. In relation to the keywords,

three groups were formed:

• Group1: (”sustainable supplier selection”,”ESG”,

”sustainability evaluation”,”assessment criteria”,)

• Group2: (”MCDM”,”sentiment analysis”, ”text

analytics”,”Artiﬁcial Intelligence”)

• Group3: (”greenwashing detection”, ”greenwash-

ing”, ”detection of eco-washing”)

2.2 Search Strategy

The search strategy combines the key concepts of our

research questions in order to retrieve accurate re-

sults. It is an organized structure of keywords, in-

cluding ”sustainable supplier selection”, ”MCDM”,

”sentiment analysis”, and ”greenwashing” as well as

ESG criteria (environmental, social, and governance)

, all related to our research questions. We then added

synonyms, variations, and related terms for each key-

word, including ESG criteria, triple bottom line cri-

teria (environmental, social, and economic) . The

use of Boolean operators (AND and OR) allows us

to explore different combinations of search terms to

improve the relevance and comprehensiveness of the

results obtained. The ﬁnal search string is (Green-

washing OR ”sustainable supplier” OR ”responsible

supplier” OR ”ESG”) AND (”sentiment analysis” OR

”MCDM” OR ”artiﬁcial intelligence” OR ”text ana-

lytics”)

2.3 Selection Criteria

After obtaining the search results from different

sources, a set of inclusion and exclusion criteria

was applied to help identify relevant primary studies.

Therefore, Inclusion Criteria (IC) are used to select

primary studies that indicate web-based analytical ap-

proaches and methods used for sustainable supplier

selection, such as MCDM , Artiﬁcial Intelligence (AI)

and sentiment analysis, or methods such as MCDM

and sentiment analysis contributing to greenwashing

detection. As for the Exclusion Criteria (EC), they

are used to eliminate those primary studies that do

not address the main topics searched in this System-

atic Literature Review (SLR), are not available, or are

directly related to an included primary study by the

same author.

• Inclusion Criteria (IC):

— Publications published in peer-reviewed jour-

nals (articles and conference papers)

— Publications published in English

— Publications published between 2019 and 2024

— Publications that are related to the research

questions

— Publications that match one of the search items

— Publications that have examples of best prac-

tices

— Publications that are related to higher education

institutions/universities

ICAART 2025 - 17th International Conference on Agents and Artiﬁcial Intelligence

452

• Exclusion Criteria (EC):

— Publications not published in peer-reviewed

journals (books and chapters)

— Publications not published in English

— Publications not published between 2019 and

2024

— Publications that are not related to the research

questions

— Publications that do not match any of the search

items

— Publications that do not have examples of best

practices

— Publications that are not related to higher edu-

cation institutions/universities

2.4 Data Collection

The search process integrates the main concepts of

our research questions to obtain precise results. The

chosen sources are listed in the table 1, along with the

associated number of papers.

Table 1: Search results by Resource.

Resource Number of papers

springer 778

IEE Xplore Digital Library 35

ACMDigital library 103

Science Direct 123

Total 1039

After ﬁltering the papers by excluding those based

on reading the abstracts, excluding more based on

reading the introductions, removing duplicate papers,

and excluding those with ﬁle not found errors, we

adopted 44 papers for this SLR. Table 2 presents the

ﬁltered search results by resources.

Table 2: Filtered search results by resources.

Resource Number of papers

springer 12

IEE Xplore Digital Library 11

ACMDigital library 0

Science Direct 21

Total reading 44

3 RESULTS AND DISCUSSION

3.1 Criteria Enumeration for

Sustainable Supplier Evaluation

To accurately evaluate sustainable suppliers, iden-

tifying comprehensive sustainability criteria is es-

sential. These criteria, encompassing environmen-

tal, social, governance (ESG), and economic factors,

form the foundation for data collection and analy-

sis. In the context of sustainable supplier evalua-

tion, the enumeration of criteria plays a crucial role

in both the Multi-Criteria Decision-Making (MCDM)

process and in Natural Language Processing (NLP)-

based sentiment analysis.

In MCDM, criteria such as carbon footprint, re-

source efﬁciency, ethical labor practices, and sup-

ply chain transparency are carefully deﬁned. Com-

mon MCDM techniques like the Analytic Hierarchy

Process (AHP) or Technique for Order of Preference

by Similarity to Ideal Solution (TOPSIS) can be em-

ployed to weigh and rank suppliers based on these cri-

teria. The effectiveness of MCDM hinges on a robust,

well-deﬁned set of criteria, as it directly inﬂuences the

decision model’s accuracy and relevance.

During the sentiment analysis phase, these sus-

tainability criteria guide the process of extracting

meaningful information from large amounts of un-

structured textual data, such as user reviews, social

media posts, or reports. Various NLP techniques can

be employed to capture stakeholder perceptions and

detect potential greenwashing.

Aspect-Based Sentiment Analysis (ABSA) allows

for granular sentiment analysis by associating speciﬁc

criteria (e.g., environmental impact, working condi-

tions) with corresponding opinions. For example,

user reviews can be parsed to identify opinions tied

to speciﬁc sustainability aspects (e.g., “The supplier

uses renewable energy, but their labor conditions are

questionable”). ABSA helps determine how stake-

holders feel about each criterion, providing insights

into the supplier’s performance on multiple fronts.

Named Entity Recognition (NER) can be used to

identify and classify named entities (e.g., organiza-

tions, materials, processes) mentioned in textual data.

In sustainable supplier selection, NER could be em-

ployed to highlight references to speciﬁc criteria such

as ”carbon emissions,” ”fair trade,” or ”energy efﬁ-

ciency,” ensuring that the analysis focuses on relevant

sustainability aspects.

Latent Dirichlet Allocation (LDA), a topic model-

ing technique, can uncover hidden topics and themes

related to sustainability in large text corpora. This

could help identify emerging sustainability concerns,

such as environmental degradation or unethical sourc-

ing practices, that may not have been previously con-

sidered in the criteria set.

Machine learning models like Support Vector Ma-

chines (SVM), Random Forests, or deep learning

models such as BERT (Bidirectional Encoder Rep-

resentations from Transformers) can classify text ac-

A Systematic Review of Sustainable Supplier Selection Using Advanced Artiﬁcial Intelligence Methods

453

cording to positive, negative, or neutral sentiments to-

ward sustainability criteria. These models are highly

effective when applied to large datasets of stakeholder

feedback, allowing the identiﬁcation of trends in per-

ception.

One of the key challenges is identifying green-

washing, where companies make exaggerated or false

claims about their sustainability efforts. Textual en-

tailment or stance detection methods can be applied

to detect inconsistencies between what companies say

(e.g., in reports or press releases) and actual stake-

holder experiences (e.g., in reviews). These tech-

niques help cross-verify the authenticity of sustain-

ability claims, ﬂagging companies that may be engag-

ing in deceptive practices.

Combining MCDM with NLP methods enhances

supplier evaluation by ensuring both quantitative

and qualitative assessments are taken into account.

MCDM methods can assign weights to different sus-

tainability criteria based on their importance, which

can then be applied in sentiment analysis to priori-

tize feedback on higher-weighted criteria (e.g., giving

more importance to environmental impact over cost).

Additionally, NLP-derived insights from stakeholder

reviews or social media can complement quantita-

tive metrics (e.g., carbon emissions, energy usage),

adding a layer of sentiment-driven decision-making

that captures public perception and trust.

The combined use of MCDM and advanced NLP

techniques allows for a thorough evaluation of sus-

tainable suppliers, ensuring transparency and reduc-

ing the risk of greenwashing. By leveraging methods

like ABSA, NER, LDA, and sentiment classiﬁcation,

organizations can extract critical insights from vast

amounts of unstructured data, aligning the evaluation

process with core sustainability criteria and improv-

ing decision-making accuracy.

Conducting a comprehensive evaluation of sus-

tainable suppliers requires more than just assessing

environmental criteria. Critical studies have under-

scored this limitation, driving many organizations to

integrate additional dimensions such as social, eco-

nomic, and governance criteria into their supplier

evaluation processes (Khan et al., 2021; Ahmadi

et al., 2020). As a result, companies increasingly

use frameworks like the Triple Bottom Line (TBL),

which encompasses economic, social, and environ-

mental dimensions (Omair et al., 2021; Rani et al.,

2020; Hoseini et al., 2020; Cheng et al., 2023; Konys,

2019; Gidiagba et al., 2023; Wang et al., 2024; Menon

and Ravi, 2022; Tavana et al., 2023), or ESG (Envi-

ronmental, Social, and Governance) criteria to ensure

a well-rounded assessment of sustainability (Khan

et al., 2021; Fischbach et al., 2024; Gupta et al., 2024;

Daying and Zi’Ao, 2023).

Incorporating Industry 4.0 criteria has become

increasingly important, aligning modern technolog-

ical advancements with sustainability goals (Tavana

et al., 2023; Fallahpour et al., 2021). Addition-

ally, resilience criteria—evaluating suppliers’ ability

to withstand and recover from disruptions such as nat-

ural disasters or economic crises—are vital. These

criteria focus on risk management, supply chain trans-

parency, and ﬂexibility, ensuring suppliers can nav-

igate unforeseen challenges while maintaining busi-

ness continuity (Ghamari et al., 2022; G

okler and Bo-

ran, 2023; Mohammed et al., 2019).

Therefore, a supplier’s sustainability should be

evaluated holistically rather than solely through eco-

logical metrics, emphasizing the need for comprehen-

sive sustainability practices. Integrating diverse cri-

teria into text analytics and data science methodolo-

gies is essential for a thorough assessment of suppli-

ers. This approach includes evaluating environmental

impacts, social responsibility, and governance prac-

tices, ensuring alignment with corporate social re-

sponsibility and ethical business standards. Adopting

such a multi-faceted evaluation approach is crucial for

leveraging information retrieval and text analytics, so-

cial analysis, and web mining to enhance the accu-

racy and transparency of sustainability assessments.

This alignment with advanced data-driven techniques

not only supports more robust supplier evaluations

but also addresses potential greenwashing, thus pro-

moting long-term business success and sustainability

goals.

The Table 3 illustrates some examples of criteria

for evaluating sustainable suppliers.

3.2 Web-Based Data Sources for

Evaluating Sustainable Suppliers

To effectively evaluate supplier sustainability, lever-

aging web-based data sources is essential for a thor-

ough and accurate assessment. Key data sources in-

clude company reports, which provide detailed in-

sights into ESG practices through annual and inte-

grated reports, adhering to international regulations

such as the Corporate Sustainability Reporting Direc-

tive (CSRD) (Twinamatsiko and Kumar, 2022). In ad-

dition, web-based data from social media platforms,

including tweets, news articles, and press releases,

offer valuable perspectives on public sentiment and

expert opinions regarding a company’s ESG perfor-

mance (Biju et al., 2023; Fischbach et al., 2024). Fur-

ther, academic and industry studies on global supply

chain risks provide context and additional informa-

tion on supplier practices at various levels (Chu et al.,

ICAART 2025 - 17th International Conference on Agents and Artiﬁcial Intelligence

454

Table 3: Criteria for evaluating sustainable suppliers.

Dimension Criteria Deﬁnition

Environmental

Wastewater Suppliers should manage wastewater to minimize pollu-

tion.

Air emissions Suppliers should reduce emissions to improve air quality.

Friendly materials Suppliers should use sustainable materials to reduce en-

vironmental impact.

Resource consumption Suppliers should conserve resources throughout produc-

tion.

Carbon emissions reduction Suppliers should strive to minimize their carbon emis-

sions to mitigate climate change impacts.

Eco-friendly packaging Suppliers should use sustainable packaging materials.

Pollution control Suppliers should prevent pollution to protect ecosystems.

Renewable electricity and

energy

Suppliers should use clean energy sources for sustainabil-

ity.

Social

Work contract Suppliers should provide stable work contracts to their

employees.

Health insurance at work Suppliers should ensure that their employees are pro-

tected by health insurance.

Economic

Quality Suppliers should strive to provide products or services

meeting agreed-upon quality standards to ensure cus-

tomer satisfaction and product sustainability.

Flexibility Suppliers should be able to adapt quickly to changes in

demand, design, or other requirements to ensure an agile

and efﬁcient supply chain.

Service Suppliers should provide responsive and quality cus-

tomer service, delivering support and appropriate solu-

tions throughout the business relationship.

Governance

Compliance Sustainable suppliers should demonstrate compliance

with relevant laws, regulations, and industry standards.

Transparency Sustainable suppliers should uphold transparency in their

operations, providing clear and accessible information.

Industry 4.0

Industry 4.0 Training and

Awareness

Suppliers should offer Industry 4.0 training and aware-

ness programs to their employees to enhance their skills,

which can contribute to better utilization of sustainable

technologies and reduced environmental impact.

Information technology (IT)

facilities

Suppliers should have adequate IT facilities, computers,

and high-speed internet access to support the adoption

and effective use of Industry 4.0 technologies, which can

promote more efﬁcient and environmentally friendly pro-

duction.

2019). By integrating these diverse web-based data

sources, organizations can achieve a more compre-

hensive and nuanced evaluation of suppliers’ sustain-

ability practices.

3.3 MCDM Methods for Evaluating

Supplier Sustainability

The evaluation of sustainable suppliers is based on a

multitude of criteria. This diversity of criteria cre-

ates a MCDM challenge, requiring a holistic evalu-

ation that considers all these dimensions. Compa-

nies are faced with the complexity of weighing and

comparing these often conﬂicting criteria to select the

most sustainable suppliers. For example, prioritizing

criteria such as environmental impact may conﬂict

with economic considerations such as cost. Corpo-

rate governance, including transparency and business

ethics, as well as corporate governance practices, are

also crucial but often difﬁcult to quantify elements.

Thus, companies must use advanced MCDM meth-

ods to integrate these criteria in a balanced way and

make informed decisions regarding the selection of

sustainable suppliers. A series of articles focuses on

the application of MCDM methods for selecting sus-

tainable suppliers (Hoseini et al., 2020; Wang et al.,

A Systematic Review of Sustainable Supplier Selection Using Advanced Artiﬁcial Intelligence Methods

455

2024; Menon and Ravi, 2022; Fallahpour et al., 2021;

okler and Boran, 2023; Yildizbasi and Arioz, 2022;

Gidiagba et al., 2023; Zhang et al., 2021; Pu

ska et al.,

2022; Khan and Ali, 2021; Masoomi et al., 2022; Liu

et al., 2019b; Chang et al., 2023; Liu et al., 2019a;

Varriale et al., 2024). Authors in (Omair et al., 2021)

introduce a decision support framework for supplier

prioritization using a MCDM approach. This frame-

work combines the Analytical Hierarchical Process

(AHP) and the Fuzzy Inference System (FIS). AHP

is employed to determine key sustainability criteria,

while FIS assesses the sustainability index of each

supplier based on these criteria. Expert opinions are

incorporated linguistically to account for the subjec-

tivity in decision-making, and fuzzy logic is used to

manage uncertainties. (Rani et al., 2020) proposes a

Pythagorean fuzzy sets (PFSs), an extension of intu-

itionistic fuzzy sets (IFSs), to manage uncertainty and

ambiguity. This study develops an approach using

PFSs and the Technique for Order Preference by Sim-

ilarity to Ideal Solution (TOPSIS) method to address

sustainable recycling partner selection problems with

unknown decision experts and criteria weights. The

study in (Tavana et al., 2023) introduces a decision

support system for evaluating and prioritizing suppli-

ers in public–private partnership projects. The ap-

proach involves two stages: ﬁrst, assessing six poten-

tial suppliers using economic, circular, social, and In-

dustry 4.0 criteria with a novel group BWM method;

and second, using fuzzy inference rules and a FIS

structure to map non-linear relationships between the

criteria and the ﬁnal score. The FIS includes 625

rules. The approach was validated using data from an

offshore wind farm project and the expertise of four

specialists. Sensitivity analysis revealed that the FIS

output is most sensitive to Industry 4.0 criteria.

The discussed works present robust methodolo-

gies for evaluating and prioritizing sustainable sup-

pliers, utilizing advanced MCDM techniques such as

AHP, FIS, and Pythagorean fuzzy sets. However,

a notable limitation across these studies is the lim-

ited consideration of user feedback in the evalua-

tion process. While these frameworks meticulously

assess suppliers based on a comprehensive set of

criteria—including environmental, social, economic,

and Industry 4.0 dimensions—they often overlook the

critical perspective provided by end-users and stake-

holders.

The MCDM approach is one of the most com-

monly used methods. However, a limitation of cur-

rent multi-criteria models for supplier selection and

performance evaluation lies in the models themselves.

Methods like the Analytic Hierarchy Process (AHP),

the Analytic Network Process (ANP), and the Fuzzy

Analytic Hierarchy Process (Fuzzy AHP) are widely

applied in most studies. These methods require

decision-makers to make judgments based on com-

parisons. While they are effective for handling vague

or qualitative information, they often restrict the num-

ber of factors and suppliers that can be analyzed si-

multaneously.

3.4 Exploring Artiﬁcial Intelligence

Approaches for Supplier

Sustainability Evaluation

The use of artiﬁcial intelligence in the evaluation of

sustainable suppliers enhances companies’ ability to

effectively implement ESG criteria in their procure-

ment processes.

AI methods excels in processing large volumes of

data (Zekhnini et al., 2023), making it particularly

suitable for analyzing vast datasets such as those re-

lated to ESG performance. Its ability to efﬁciently

process large amounts of information and identify

patterns and trends within the data signiﬁcantly en-

hances its effectiveness in evaluating supplier sus-

tainability. This capability to handle big data sets

AI apart as a powerful tool for sustainability assess-

ment and decision-making in procurement processes.

In the context of supplier evaluation, artiﬁcial intelli-

gence methods are divided into three categories: Ma-

chine Learning (ML), Deep Learning (DL), and Nat-

ural Language Processing (NLP). Each category has

distinct applications.

3.4.1 Machine Learning Approaches

ML has recently attracted more attention as a result of

its effectiveness in a variety of applications ranging

from image categorization to a variety of decision-

making challenges. The applications of ML ap-

proaches have increased dramatically in recent years

due to the explosion of data. In a predictive setting,

methods such as data envelopment analysis super-

vised learning, and unsupervised learning has shown

very outstanding performance. Furthermore, machine

learning can tolerate inaccuracies, uncertainty, and

imprecise information to achieve robustness when

replicating human decision-making behavior. These

functions not only solve the problem of scalability and

rapidity, but they also reduce the drawbacks of ear-

lier approaches and meet the demands of ever more

difﬁcult supplier networks (Jagyasi and Raut, 2023).

In (Kulkarni et al., 2023; Baqi et al., 2022), authors

emphasize using AI to optimize the analysis of sup-

pliers’ ESG data, thereby facilitating ESG perfor-

mance monitoring and measurement. Additionally,

ICAART 2025 - 17th International Conference on Agents and Artiﬁcial Intelligence

456

(Ahmad et al., 2023) proposes using ML to analyze

ESG data in the context of sustainability, with a fo-

cus on auto mating data collection and analysis. Fur-

thermore, in the context of evaluating sustainable sup-

pliers, employing ML techniques, particularly regres-

sion analysis, could be considered to analyze ESG

data and identify correlations with company perfor-

mance, thereby contributing to more responsible in-

vestment decisions (Twinamatsiko and Kumar, 2022).

Concurrently, AI can also play a vital role in re-

ducing greenhouse gas emissions by enabling precise

data analysis and identifying areas where companies

can enhance their environmental sustainability (Gaur

et al., 2023).

These various approaches illustrate the diversity

of methods and applications of AI in assessing sus-

tainable suppliers, offering unique perspectives on

how this technology can promote sustainability within

businesses.

The results of the experiment demonstrated the

effectiveness of this approach, with most classiﬁca-

tion algorithms achieving an accuracy of over 90%.

Speciﬁcally, models such as LGBM, extra tree, gradi-

ent boosting, and decision tree exceeded expectations

by achieving an accuracy of over 99%.

(Abdulla et al., 2019) proposes a hybrid model for

supplier selection by integrating AHP with a machine

learning model. They utilized a decision tree classiﬁer

to distinguish between good and bad suppliers.

ML methods are used for predicting ﬁnancial per-

formance from ESG data, as well as for calculat-

ing ESG scores from ﬁnancial data, where experi-

ments have shown that ML algorithms, particularly

gradient boosting and XGB, yielded accurate predic-

tions. Furthermore, anomaly detection techniques

based on ML, particularly the Local Outlier Factors

(LOF) model, can be deployed to identify outliers

or unusual values in ESG datasets, thus providing a

robust methodology for detecting any noise or mali-

cious data manipulation that could compromise their

integrity and reliability (Lee et al., 2022)

3.4.2 Deep Learning Approaches

In recent years, various deep learning methods have

been applied to the ﬁeld of sustainable supplier se-

lection, leveraging their ability to handle complex,

high-dimensional data and capture intricate patterns

(Nicherala et al., 2022).. Techniques such as Convo-

lutional Neural Networks (CNNs) have been used for

their strength in feature extraction and pattern recog-

nition, especially when dealing with large datasets

that include images or spatial data. Recurrent Neural

Networks (RNNs) and their variants, like Long Short-

Term Memory (LSTM) networks, are employed to

analyze sequential data, such as time-series informa-

tion related to supplier performance. Autoencoders

have been utilized for dimensionality reduction and

anomaly detection, helping to identify outliers and

inconsistencies in supplier data. Furthermore, hy-

brid models combining deep learning with traditional

methods, like Decision Trees or the Analytic Hier-

archy Process (AHP), have been developed to en-

hance decision-making by integrating structured de-

cision frameworks with the powerful data processing

capabilities of deep learning.

Another recent study in the realm of sustainabil-

ity and ESG data utilizes advanced DL techniques to

analyze ESG data. In this experiment, deep learn-

ing techniques were employed to classify news arti-

cles into ESG labels. Speciﬁcally, embedding and

Bi-LSTM (Bidirectional Long Short-Term Memory)

models were utilized for this multi-class classiﬁcation

task. The results indicated signiﬁcant performance

improvements when employing these deep learning

techniques, achieving a maximum prediction score of

0.8582 without processing Stop Words and 0.8936 af-

ter processing Stop Words. This demonstrates the

effectiveness of deep learning in accurately catego-

rizing news articles based on their ESG attributes.

Additionally, the experiment highlighted the impor-

tance for institutions to promptly address ESG is-

sues to safeguard their reputation and value, given

the rapid dissemination of company-related informa-

tion through the media. Therefore, the utilization of

deep learning models like Bi-LSTM showcases their

capability to efﬁciently analyze and categorize news

articles, providing valuable insights into companies’

ESG activities and facilitating timely responses to

emerging issues.(Lee et al., 2022)

3.4.3 NLP and Web-Driven Sentiment Analysis

Approaches

Natural Language Processing (NLP) is a branch of ar-

tiﬁcial intelligence that deals with language process-

ing and knowledge extraction. There are multiple ap-

plications of NLP, from sentiment analysis to key-

word extraction, text classiﬁcation, and text summa-

rization. Since NLP plays a crucial role in text anal-

ysis, advanced techniques such as BERT and YAKE

are used to explore and understand sustainability re-

ports, particularly focusing on ESG aspects. BERT is

a pre-trained language model that supports tokeniza-

tion. It uses a tokenization technique called ”Word-

Piece,” which breaks down words into smaller sub-

units called ”subwords” for a more efﬁcient and ﬂex-

ible representation of language. However YAKE ex-

cels at keyword extraction by automatically analyz-

ing a text to identify the most important terms espe-

A Systematic Review of Sustainable Supplier Selection Using Advanced Artiﬁcial Intelligence Methods

457

cially related to sustainability criteria and values. A

data preprocessing phase is prominent after data col-

lection, where BERT tokenization can be employed.

This involved subdividing the data into keywords,

with ESG subcategories associated with the extracted

keywords and identiﬁed sustainability criteria (Gupta

et al., 2024). The ESG-Miner is a sophisticated tool

designed to evaluate companies’ ESG performance

through the analysis of news headlines. It begins

by automatically detecting speciﬁc companies men-

tioned in headlines using named entity recognition

and string matching with TF-IDF and cosine similar-

ity. Then, it classiﬁes these headlines in two steps:

ﬁrst, by determining their ESG relevance using ML

and DL models such as BERT, TF-IDF, and SVM;

second, by assigning relevant headlines to one of the

three ESG categories. For sentiment analysis, the

tool identiﬁes whether the company’s behavior men-

tioned is perceived as neutral, positive, or negative,

using a classiﬁer trained on a manually annotated

corpus (Fischbach et al., 2024). An other study fo-

cuses on the selection of regions in the global supply

chain, taking regional differences into account. While

the global supply chain offers advantages such as in-

creased ﬂexibility and cost reduction, it also presents

various risks depending on the regions. The objec-

tive of this research is to identify these regional risks

through an analysis based on text mining. To achieve

this, a corpus of 11 well-cited and relevant academic

articles on global supply chain risk management was

compiled. These articles were imported into RStudio

for text data preprocessing using the tm and tidytext

packages. Additionally, the words were transformed

into their stems using the SnowballC package. Au-

thors in (Chu et al., 2019) focus on measuring the

ESG impacts in African cities using topic-based sen-

timent analysis methodologies applied to datasets col-

lected from social media platforms. The aim is to un-

derstand the population’s perception of ESG impacts,

given their signiﬁcant inﬂuence on society. The pro-

cess begins with Data Collection: The study utilizes

basic keywords from the 40 Cities framework, cover-

ing themes aligned with the Sustainable Development

Goals (SDGs). Data is collected from social media

platforms like Twitter using Snscrape, and from web

pages via Google Search using Selenium and Traﬁ-

latura. Then, in the Data Filtering phase: Tweets con-

taining both city names and keywords are considered

relevant.

3.5 Greenwashing Avoidance Using

Sentiment Analysis

Despite the effectiveness of mentioned methods and

approaches instead of objective evaluation methods

based on criteria for sustainable suppliers, these eval-

uations remain insufﬁcient as it does not account for

the risk of greenwashing. Greenwashing is ”the dis-

semination of false or deceptive information regard-

ing an organization’s environmental strategies, goals,

motivations, and actions”(Aronczyk et al., 2024),re-

ferring to the practice of misleadingly or exaggerat-

edly presenting a company’s practices or products as

environmentally friendly when they may not neces-

sarily be so.

Sentiment analysis plays a crucial role in detect-

ing and avoiding greenwashing by enabling a critical

evaluation of corporate communications. In the case

of Pathways Alliance, this method helps identify indi-

cators of greenwashing by analyzing the tone and per-

ception of public messages, thereby revealing poten-

tial discrepancies between sustainability promises and

actual actions. (Aronczyk et al., 2024). Another study

shows that sentiment analysis is conducted using a

qualitative approach, particularly analyzing Twitter

data related to ”ESG” and ”greenwashing”. The

methodology relies on the use of MAXQDA software,

which provides a reliable solution for analyzing qual-

itative data, including sentiments expressed in tweets.

To perform this analysis, researchers extracted tweets

containing the keywords ”ESG” and ”greenwashing”

separately. Subsequently, they utilized MAXQDA

software to analyze this data and assess the sentiments

expressed in the tweets. MAXQDA evaluates senti-

ments by assigning a sentiment score to each word in

the lexicon used, categorizing words in terms of pos-

itive, neutral, or negative connotations. Tweets are

then automatically coded with sentiment labels, en-

abling rapid capture of public sentiment on a speciﬁc

research topic. The results of this sentiment analy-

sis revealed a positive correlation between sentiments

toward ESG and greenwashing. This indicates that

concerns regarding greenwashing have increased pro-

portionally with interest in ESG-related issues (Biju

et al., 2023).

4 CONCLUSIONS

In this SLR, we have explored the growing trend

of sustainable supplier evaluation in today’s business

landscape, where sustainability has become a ma-

jor concern. However, evaluating the sustainability

of suppliers poses a complex challenge that requires

ICAART 2025 - 17th International Conference on Agents and Artiﬁcial Intelligence

458

a holistic approach, considering various criteria and

data sources. Furthermore, we have discussed various

evaluation methods available, including multi-criteria

decision-making (MCDM) and artiﬁcial intelligence

approaches such as machine learning and deep learn-

ing. These methods enable a comprehensive analy-

sis of supplier performance across a diverse range of

criteria, thereby facilitating informed and data-driven

decision-making.

While many traditional methods such as multi-

criteria decision-making (MCDM), machine learning,

and deep learning are used for this evaluation, they

present potential risks of greenwashing and lack of

transparency. It is important to note that most existing

studies primarily focus on evaluating sustainable sup-

pliers based on expert opinions in the ﬁeld. However,

this approach may be limited as it does not always

consider the opinions and sentiments of end-users and

public feedback, which can offer important perspec-

tives on the actual sustainability of suppliers.

With this in mind, a promising future approach

would be to integrate user feedback analysis into the

process of evaluating sustainable suppliers. By ex-

amining user feedback on supplier services, valuable

insights could be obtained into their performance in

terms of sustainability perceived by consumers. This

approach would complement traditional evaluations

based on expert opinions with data from direct and au-

thentic sources, thus providing a more comprehensive

and balanced perspective on supplier sustainability.

REFERENCES

Abdulla, A., Baryannis, G., and Badi, I. (2019). Weight-

ing the key features affecting supplier selection using

machine learning techniques.

Ahmad, V., Goyal, L., Arora, M., Kumar, R., Chythanya,

K. R., and Chaudhary, S. (2023). The impact of ai

on sustainability reporting in accounting. In 2023 6th

International Conference on Contemporary Comput-

ing and Informatics (IC3I), volume 6, pages 643–648.

IEEE.

Ahmadi, H. B., Lo, H.-W., Gupta, H., Kusi-Sarpong, S.,

and Liou, J. J. (2020). An integrated model for select-

ing suppliers on the basis of sustainability innovation.

Journal of Cleaner Production, 277:123261.

Aronczyk, M., McCurdy, P., and Russill, C. (2024). Green-

washing, net-zero, and the oil sands in canada: The

case of pathways alliance. Energy Research and So-

cial Science, 112:103502.

Baqi, A., Abdeldayem, M. M., and Aldulaimi, S. H. (2022).

Embedding artiﬁcial intelligence and green ideology

in formulating corporate and marketing strategies.

In 2022 ASU International Conference in Emerging

Technologies for Sustainability and Intelligent Sys-

tems (ICETSIS), pages 1–4.

Biju, A., Kodiyatt, S., and Krishna, P. e. a. (2023). Esg sen-

timents and divergent esg scores: suggesting a frame-

work for esg rating. In SN Business and Economics,

volume 3, page 209.

Chang, J.-P., Chen, Z.-S., Wang, X.-J., Mart

ınez,

L., Pedrycz, W., and Skibniewski, M. J. (2023).

Requirement-driven sustainable supplier selection:

Creating an integrated perspective with stakeholders’

interests and the wisdom of expert crowds. Computers

and Industrial Engineering, 175:108903.

Cheng, C., Wang, X., and Ren, X. (2023). Selection of

outsourcing logistics providers in the context of low-

carbon strategies. Environmental Science and Pollu-

tion Research, 30(7):18701–18717.

Chu, C.-Y., Park, K., and Kremer, G. E. (2019). Apply-

ing text-mining techniques to global supply chain re-

gion selection: Considering regional differences. Pro-

cedia Manufacturing, 39:1691–1698. 25th Interna-

tional Conference on Production Research Manufac-

turing Innovation: Cyber Physical Manufacturing Au-

gust 9-14, 2019 — Chicago, Illinois (USA).

Daying, Y. and Zi’Ao, Y. (2023). Discovering variation ﬁ-

nancial performance of esg scoring through big data

analytics. In 2023 Asia-Europe Conference on Elec-

tronics, Data Processing and Informatics (ACEDPI),

pages 141–150.

Fallahpour, A., Wong, K. Y., Rajoo, S., Fathollahi-Fard,

A. M., Antucheviciene, J., and Nayeri, S. (2021). An

integrated approach for a sustainable supplier selec-

tion based on industry 4.0 concept. Environmental

science and pollution research, pages 1–19.

Fischbach, J., Adam, M., Dzhagatspanyan, V., Mendez, D.,

Frattini, J., Kosenkov, O., and Elahidoost, P. (2024).

Automatic esg assessment of companies by mining

and evaluating media coverage data: Nlp approach

and tool.

Gaur, L., Afaq, A., Arora, G. K., and Khan, N. (2023). Ar-

tiﬁcial intelligence for carbon emissions using system

of systems theory. Ecological Informatics, 76:102165.

Ghamari, R., Mahdavi-Mazdeh, M., and Ghannadpour, S. F.

(2022). Resilient and sustainable supplier selection

via a new framework: a case study from the steel in-

dustry. Environment, development and sustainability,

pages 1–39.

Gidiagba, J., Tartibu, L., and Okwu, M. (2023). Sustainable

supplier selection in the oil and gas industry: An inte-

grated multi-criteria decision making approach. Pro-

cedia Computer Science, 217:1243–1255. 4th Inter-

national Conference on Industry 4.0 and Smart Man-

ufacturing.

Gupta, A., Chadha, A., and Tewari, V. (2024). A natural

language processing model on bert and yake technique

for keyword extraction on sustainability reports. IEEE

Access, 12:7942–7951.

okler, S. H. and Boran, S. (2023). A novel resilient and

sustainable supplier selection model based on d-ahp

and dematel methods. Journal of Engineering Re-

search.

Hoseini, A. R., Ghannadpour, S. F., and Ghamari, R.

(2020). Sustainable supplier selection by a new

A Systematic Review of Sustainable Supplier Selection Using Advanced Artiﬁcial Intelligence Methods

459

possibilistic hierarchical model in the context of z-

information. Journal of Ambient Intelligence and Hu-

manized Computing, 11:4827–4853.

Jagyasi, D. and Raut, A. R. (2023). Implementation of esg

index on long-term value and performance of ogani-

zations using ai and ml. In 2022 OPJU International

Technology Conference on Emerging Technologies for

Sustainable Development (OTCON), pages 1–5.

Khan, A. U. and Ali, Y. (2021). Sustainable supplier selec-

tion for the cold supply chain (csc) in the context of

a developing country. Environment, development and

sustainability, pages 1–30.

Khan, S. A. R., Yu, Z., Golpira, H., Sharif, A., and Mardani,

A. (2021). A state-of-the-art review and meta-analysis

on sustainable supply chain management: Future re-

search directions. Journal of Cleaner Production,

278:123357.

Konys, A. (2019). Methods supporting supplier selec-

tion processes – knowledge-based approach. Proce-

dia Computer Science, 159:1629–1641. Knowledge-

Based and Intelligent Information and Engineering

Systems: Proceedings of the 23rd International Con-

ference KES2019.

Kulkarni, A., Joseph, S., and Patil, K. (2023). Role of artiﬁ-

cial intelligence in sustainability reporting by leverag-

ing esg theory into action. In 2023 International Con-

ference on Advancement in Computation and Com-

puter Technologies (InCACCT), pages 795–800.

Lee, O., Joo, H., Choi, H., and Cheon, M. (2022). Propos-

ing an integrated approach to analyzing esg data via

machine learning and deep learning algorithms. Sus-

tainability, 14(14).

Liu, A., Xiao, Y., Lu, H., Tsai, S.-B., and Song, W. (2019a).

A fuzzy three-stage multi-attribute decision-making

approach based on customer needs for sustainable

supplier selection. Journal of Cleaner Production,

239:118043.

Liu, H.-C., Quan, M.-Y., Li, Z., and Wang, Z.-L. (2019b).

A new integrated mcdm model for sustainable sup-

plier selection under interval-valued intuitionistic un-

certain linguistic environment. Information Sciences,

486:254–270.

Masoomi, B., Sahebi, I. G., Fathi, M., Yıldırım, F., and

Ghorbani, S. (2022). Strategic supplier selection for

renewable energy supply chain under green capabil-

ities (fuzzy bwm-waspas-copras approach). Energy

Strategy Reviews, 40:100815.

Menon, R. R. and Ravi, V. (2022). Using ahp-topsis

methodologies in the selection of sustainable suppli-

ers in an electronics supply chain. Cleaner Materials,

5:100130.

Mohammed, A., Harris, I., Soroka, A., and Nujoom, R.

(2019). A hybrid mcdm-fuzzy multi-objective pro-

gramming approach for a g-resilient supply chain net-

work design. Computers and Industrial Engineering,

127:297–312.

Nicherala, Y. K., Sadula, S., and Shrinivas, V. P. (2022).

Deep learning based sustainable material attribution

for apparels. In 2022 IEEE 18th International Confer-

ence on Automation Science and Engineering (CASE),

pages 1352–1357.

Omair, M., Noor, S., Tayyab, M., Maqsood, S., Ahmed,

W., Sarkar, B., and Habib, M. (2021). The selection

of the sustainable suppliers by the development of a

decision support framework based on analytical hier-

archical process and fuzzy inference system. Inter-

national Journal of Fuzzy Systems, 23(7):1986–2003.

Association.

ska, A., Beganovi

c, A., Stojanovi

c, I., and Murti

c, S.

(2022). Green supplier’s selection using economic and

environmental criteria in medical industry. Environ-

ment, Development and Sustainability, pages 1–22.

Rani, P., Mishra, A. R., Rezaei, G., Liao, H., and Mardani,

A. (2020). Extended pythagorean fuzzy topsis method

based on similarity measure for sustainable recycling

partner selection. International Journal of Fuzzy Sys-

tems, 22:735–747.

Tavana, M., Sorooshian, S., and Mina, H. (2023). An inte-

grated group fuzzy inference and best–worst method

for supplier selection in intelligent circular supply

chains. Annals of Operations Research, pages 1–42.

Twinamatsiko, E. and Kumar, D. (2022). Incorporating esg

in decision making for responsible and sustainable in-

vestments using machine learning. In 2022 Interna-

tional Conference on Electronics and Renewable Sys-

tems (ICEARS), pages 1328–1334.

Varriale, V., Cammarano, A., Michelino, F., and Caputo,

M. (2024). The role of digital technologies in pro-

duction systems for achieving sustainable develop-

ment goals. Sustainable Production and Consump-

tion, 47:87–104.

Vinella, A., Capetz, M., Pattichis, R., Chance, C., and

Ghosh, R. (2023). Leveraging language models to de-

tect greenwashing.

Wang, Y., Wang, W., Wang, Z., Deveci, M., Roy, S. K., and

Kadry, S. (2024). Selection of sustainable food suppli-

ers using the pythagorean fuzzy critic-marcos method.

Information Sciences, 664:120326.

Yildizbasi, A. and Arioz, Y. (2022). Green supplier

selection in new era for sustainability: A novel

method for integrating big data analytics and a hybrid

fuzzy multi-criteria decision making. Soft Comput.,

26(1):253–270.

Zekhnini, K., Chaouni Benabdellah, A., and Cherraﬁ, A.

(2023). A multi-agent based big data analytics sys-

tem for viable supplier selection. Journal of Intelli-

gent Manufacturing, pages 1–21.

Zhang, J., Li, L., Zhang, J., Chen, L., and Chen, G. (2021).

Private-label sustainable supplier selection using a

fuzzy entropy-vikor-based approach. Complex and In-

telligent Systems, pages 1–18.

ICAART 2025 - 17th International Conference on Agents and Artiﬁcial Intelligence

460