FASTER-AI: A Comprehensive Framework for Enhancing the

Trustworthiness of Artificial Intelligence in Web Information

Systems

Christos Troussas

, Christos Papakostas

, Akrivi Krouska

, Phivos Mylonas

and Cleo Sgouropoulou

Department of Informatics and Computer Engineering, University of West Attica, Egaleo, Greece

Keywords: Trustworthy AI, Web Information Systems, Fairness and Bias Mitigation, Explainability and Transparency,

Security and Privacy, Robustness, Ethical AI, AI Implementation Framework, AI Evaluation Roadmap, Case

Studies in AI, FASTER-AI.

Abstract: With increasing embedding of artificial intelligence (AI) in web information systems (WIS), the maximum

assurance on the reliability of such AI systems is solicited. Although this aspect is gaining importance, no

comprehensive framework has yet been developed to ensure AI reliability. This paper aims to bridge that gap

by proposing the AI FASTER framework to enhance the reliability of AI in WIS. The key dimensions of

concern within the framework are FASTER-AI: Fairness/bias mitigation, explainability/transparency,

security/privacy, robustness, and ethical considerations/accountability. Each one guides in precisely the area

where trust shall be accomplished: a decrease in bias, model interpretability, protection of data, resilience of

models, and ethics in governance. The implementation methodology for these dimensions involves

preliminary assessment, planning, integration, testing, and continuous improvement. Validation of proof for

FASTER-AI was created based on in-depth case studies across different verticals: e-commerce, finance,

health, and fraud detection. This work has demonstrated how FASTER-AI is applied through illustrative case

studies showing promising performance. From the initial results of high improvement in terms of fairness,

transparency, security, and robustness, it may be effectively inferred that FASTER-AI can be successfully

applied.

1 INTRODUCTION

Web Information Systems (WIS) have grown to

become the backbone of the digital space, without

which the management, storage, and delivery of

information over the internet would not be possible.

Applications range from e-commerce websites, social

media networks, online education, content

management systems, etc., basing on seamless data

flow and user interaction (Bhutani & Mittal, 2023; Ge

et al., 2023; Huang, 2022; Kardaras et al., 2024;

Troussas et al., 2015, 2022; Virvou et al., 2012;

Zenkert & Fathi, 2023). Efficiency and reliability

https://orcid.org/0000-0002-9604-2015

https://orcid.org/0000-0002-5157-347X

https://orcid.org/0000-0002-8620-5255

https://orcid.org/0000-0002-6916-3129

https://orcid.org/0000-0001-8173-2622

become crucial aspects of WIS, being that such issues

relate to the experience of users, data accessibility,

and overall functionality important to billions of

people based on daily internet service.

Artificial Intelligence (AI) being integrated into

WIS, does bring a new change in the internal

operation methodology. WIS can be extended

through AI technologies for improved ability to

automate complex tasks and personalize user

experience more efficiently in the optimization of

content delivery methods (Ganesh & Rastogi, 2023;

Krouska et al., 2020). Examples include

recommendation systems, e-commerce with the

power of AI-driven algorithms in e-commerce,

Troussas, C., Papakostas, C., Krouska, A., Mylonas, P. and Sgouropoulou, C.

FASTER-AI: A Comprehensive Framework for Enhancing the Trustworthiness of Artiﬁcial Intelligence in Web Information Systems.

DOI: 10.5220/0013061100003825

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

In Proceedings of the 20th International Conference on Web Information Systems and Technologies (WEBIST 2024), pages 385-392

ISBN: 978-989-758-718-4; ISSN: 2184-3252

385

content filtering on social media, and automated

customer support on various online services. It

enables a system to improve operating efficiency and,

more importantly, fosters innovation in driving the

empowerment of Web information systems to offer

more intelligent, responsive services.

In particular, modern web information systems

are increasingly dependent on AI, further

complicating the critical concern of trustworthiness.

Fairness, transparency, security, robustness, and

ethics were assumed within the frame of the AI when

the AI systems take decisions based on inputs to

directly affect the user's decisions, for instance,

product recommendations, content moderation, or

management of private sensitive personal data in

(Martin, 2022). Only truthful AI can be maintained

for user trust, and it will also go on to serve in meeting

regulatory requirements to protect against such harms

due to biased decision-making, data breach, or actions

which are not ethically right. Without a solid

foundation of trust, the potential benefits arising from

AI within web information systems cannot be realized

fully, mainly since the associated risks in decisions

taken by AI may also crush the integrity and

credibility of such systems.

The literature has matured well on the issue of AI

trustworthiness, with key dimensions such as

fairness, explainability, security, robustness, and

ethics. For example, Mehrabi et al. (2021) discussed

AI bias and fairness, which proposed methods for bias

detection and mitigation. For instance, in the aspect

of explainability, some of these advance to better

steps of constructing the AI-interpretable models, like

LIME and SHAP when announced in 2016 by Ribeiro

et al. It is a security loophole, as described in the

research topic on adversarial attacks and defenses by

Carlini et al., 2019. Another proposition to harden AI

models was made by Goodfellow et al. in 2015 and

later modified by Madry et al. in 2017, but this

adversarial training has remained challenging for

real-time application in web information systems.

Several frameworks do attempt at ensuring

trustworthiness of AI but are scoped only to narrow

levels. Floridi et al. (2018) developed an ethical

framework for AI and robotics; however, it is devoid

of concrete technical guidelines for WIS. Shahriari &

Shahriari (2017) submitted ethical guidelines for AI

development; however, these omitted the issues that

are specific to WIS. Rieke et al. (2020) proposed a

GDPR-compliant, privacy-preserving AI framework

dealing largely with data privacy aspects.

The earlier works, while useful, often address

individual trust dimensions such as fairness in

Agarwal et al. (2018) or transparency in Doshi-Velez

and Kim (2017) without placing them into context

with respect to WIS. Works such as Veale & Binns

2017 and Danks & London 2017 discuss fairness,

accountability, and social impact brought about by AI

but do not go far enough to offer a more holistic and

multi-dimensional framework aimed at WIS. Yet

other proposals articulate high-level ethical

considerations but do not incorporate within the

proposals themselves low-level technical safeguards,

such as security and robustness specific to WIS.

In turn, corollary to the pervasiveness of AI in

web information systems today, there is thus a

pressing need for an integrated framework

surrounding concerns across all facets of trustworthy

AI in such settings. The paper proposes the FASTER-

AI framework related to five main dimensions:

Fairness and Bias Mitigation, Explainability and

Transparency, Security and Privacy, Robustness, and

Ethical Considerations and Accountability.

FASTER-AI has been designed to populate practical

guidelines for the organizations and developers on

how to make their AI systems in WIS more

trustworthy. The framework provided by FASTER-

AI can be implemented to construct AI-driven WIS

whose performance is optimal using credible models,

and to ensure the best achievable level of trust and

dependability.

2 FASTER-AI FRAMEWORK

FOR TRUSTWORTHY AI IN

WIS

Although significant breakthroughs have been

realized in fast-tracking AI integration into the WIS

and automation of jobs for improved user experiences

and processes, at the same time, it tends to lead to

several concerns about dependability of AI systems

related to fairness, transparency, security, robustness,

and ethics. In view of the above challenges, we

propose a framework built upon insights and

approaches detailed in the literature presented in

Section 2 that could enable further extension of work

on development and evaluation of trustworthy AI in

WIS. Figure 1 presents the proposed framework over

five dimensions: Fairness and Bias Mitigation;

Explainability and Transparency; Security and

Privacy; Robustness; and Accountability and Ethics.

These start with the first letter of the dimensions:

Fairness, Accountability, Security, Transparency,

Ethical considerations, Robustness-adding AI to

close off to the acronym with its full meaning

underlining, as said, the focus on fostering

WEBIST 2024 - 20th International Conference on Web Information Systems and Technologies

386

trustworthiness within a given context, in particular

that of artificial intelligence systems. Each of these

dimensions contributes much to the overall credibility

of the AI system and, when integrated, provides

pragmatic frameworks for organizations to

operationalize AI in online contexts.

Figure 1: FASTER-AI Structure.

The first dimension, Fairness and Bias Mitigation,

is about the need for the AI systems to avoid

discriminatory outcomes. For WIS, where choices

made by AI could involve a diverse user base,

fairness will be critical in maintaining public trust and

meeting legal obligations. FASTER-AI emphasizes

that ongoing surveillance and evaluation of AI

models have to be performed in order to continuously

identify biases and take action against them

accordingly. Some of the strategies ensuring the

provision of justice to all groups of users could

include fairness-aware learning algorithms, re-

weighting the training datasets, and modifications in

post-processing. Furthermore, fairness audits should

be built into the AI lifecycle of development to

continually evaluate and update models with ever-

exposed data and scenarios.

The second dimension refers to making clear and

explaining how artificial intelligence systems

improve the understandability of the latter for the

end-users and stakeholders. In the context of WIS,

such decisions are likely to have far-reaching

repercussions on users, with the transparency and

interpretability of AI in such decisions becoming

paramount. In general, FASTER-AI encourages the

use of interpretable models or tools that explain how

the black-box models work, such as Local

Interpretable Model-agnostic Explanations (LIME)

and SHapley Additive exPlanations (SHAP). On the

other hand, explanation interfaces should be designed

in a user-centered way because they are insight-

presenting interfaces that transform intelligible and

understandable forms to the needs of technical

experts, end-users, and regulatory entities. Faster-AI

makes its software sold transparently, hence gaining

user trust and propping more informed decision-

making for all stakeholders.

The third dimension of our framework is Security

and Privacy. This imperative relates to protection

from AI systems against intrusion of the destructive

nature that may affect user data confidentiality. Since

the information being handled in WIS is sensitive,

strict security mechanisms need to be implemented to

ensure security. Because of this, FASTER-AI

encourages adversarial training for safeguarding AI

models from bad inputs performing privacy-

preserving operations such as differential privacy,

federated learning, and homomorphic encryption. It

accomplishes these through methods that ensure AI

systems operate securely and privately in

decentralized settings or where there is a high risk of

breaching the data. Further, FASTER-AI identifies

encryption mechanisms and access controls that

impede illegitimate persons from accessing sensitive

information, as well as compliance with the

concerned protection regulations such as GDPR or

CCPA.

Robustness, being the fourth dimension, aims at

assuring an AI system's reliability and effectiveness,

regardless of the range of conditions in WIS. These

models should be robust to distributional shifts,

outliers, and adversarial circumstances that could

render them unpredictably harmful. FASTER-AI

advocates for increased testing on edge cases, which

the methodologies of cross-domain generalization

and transfer learning can help with. Similarly, the

constant monitoring and retraining of artificial

intelligence models to adapt them to shifting data

distributions and operational environments are

recommended so that such systems remain robust and

trustworthy throughout their life cycle.

The fifth dimension, Ethical Considerations and

Accountability, is a very important parameter with

respect to the integrity and social acceptability of AI

systems. FASTER-AI embeds ethical considerations

such as fairness, justice, and respect for individual

autonomy into the design and operation of AI models.

It also places great emphasis on accountability

structures such as full documentation, audit trails, and

organizational oversight bodies. Such initiatives

ensure that artificial intelligence systems are in

bounds of fixed parameters of ethics, and any

problem or non-conformity may quickly be brought

under the control of the concerned authority.

FASTER-AI: A Comprehensive Framework for Enhancing the Trustworthiness of Artiﬁcial Intelligence in Web Information Systems

387

Additionally, FASTER-AI also recommends the

artificial intelligence be governed by frameworks

created through ethics committees, which can

continuously provide oversight and assurance that AI

systems conform to the moral values of the

communities they serve.

Faster-AI embeds multiple dimensions of AI

trustworthiness, including fairness, transparency,

security, and robustness, into one framework that is

important for WIS. It also underlines the fact that an

end which is expected to be fair needs transparency

and robust security protocols relevant to privacy

protection. With this approach, it is possible for

organizations to ensure state-of-the-art standards for

AI systems while at the same time aligning with

citizen values.

With such a wide variety of users and dynamic

data, WIS introduces a very special set of challenges

into the world of AI systems, which should be

proficient at decision-making and trustworthy.

FASTER-AI addresses this issue by developing

appropriate metrics for each of these dimensions-

fairness, transparency, and security-which permit

periodic assessments and their alignment with ethical

standards. This framework should be agile and

adaptable to the needs in various sectors, like

healthcare or e-commerce, and scalable for

organizations. In this regard, FASTER-AI will enable

organizations to build trust and credibility in their AI

systems by being compliant with regulations and

societal expectations.

The major steps of the FASTER-AI

implementation methodology include: an assessment

phase that audits the existing AI systems for their

fairness, transparency, and security by aligning with

the dimensions of FASTER-AI. The planning and

design phase is next, during which the tools required,

AI Fairness 360 and SHAP, would be identified in a

manner to ensure the involvement of stakeholders.

Next comes the integration phase, where cross-

functional teams—data scientists and lawyers—come

together to apply these tools. A wide range of datasets

and scenarios during the testing and validation phase

ensure that AI models are nondiscriminatory, robust,

and transparent. In this iteration process, users

perform ongoing monitoring and improvement

through KPIs and real-time dashboards; issues that

emerge from it are identified and fixed.

The last step involves reporting and

communication, through which updates would

continue to come in via reports, audits, and

communication strategies, thereby building trust in

users for the long term about the technical and ethical

standards followed.

3 ROADMAP FOR EVALUATION

AND PRELIMINARY RESULTS

In the operation to assess how well FASTER-AI is

able to enhance the trustworthiness of AI systems in

WIS, there needs to be an efficient evaluation strategy

in place. This section presents a general evaluation

framework for FASTER-AI and the central

dimensions: fairness, transparency, security,

robustness, and ethical accountability. Finally, the

interim outcomes of some first case studies are

analyzed to provide preliminary views concerning the

practical impact of FASTER-AI.

3.1 Evaluation Roadmap

However, the value of FASTER-AI has to be tested

within several real-world settings where the AI

systems are crucial. It measures best the impact

caused by the implementation of FASTER-AI within

each one of these dimensions.

Fairness is evident in the AI-driven decisions

when they affect people from different demographic

groups. For example, the fairness evaluation of an e-

commerce recommendation system can be done

based on the distribution of recommendations in

different user segments, using Disparate Impact Ratio

and Demographic Parity metrics. It shall be able to

demonstrate if FASTER-AI efficiently removed

biases and promoted fairness.

The transparency domain in financial services will

be a focal point in this research where the

interpretability of the decision by AI can either be

important or very important. This will consist of

deploying mechanisms that provide transparency

behind the AI decisions and improve the AI decisions'

interpretability to end-users. Associated domains

with such measures may include customer

satisfaction ratings, disputes, regulatory compliance,

among other indicators of performance.

This would imply security and privacy tests under

environments dealing with sensitive information; say,

healthcare institutions. In this case, evaluation will be

done on how much potential FASTER-AI has in

protecting data, factoring in protection measures,

including encryption, differential privacy, and

federated learning. This would have been evaluated

by the number of data breaches, testing robustness of

the security protocols against simulated attacks, and

also gauging data privacy stakeholder confidence.

Robustness will be studied through controlled

experiments that will expose the AI models to

adversarial conditions, such as data corruption or

malicious attacks. Comparison of error rate and

WEBIST 2024 - 20th International Conference on Web Information Systems and Technologies

388

resilience of the models before and after applying the

strategies for improving robustness recommended by

FASTER-AI will enable insight into the framework's

capacity for improving stability and reliability under

stressful conditions.

Governance frameworks will help organizational

ethical accountability with the constitution of AI

ethics boards. The effectiveness of governance

frameworks will be measured on the extent to which

they solve ethical dilemmas, introduce transparency

in the decision-making process, and maintain

compliance with set ethical standards.

3.2 Preliminary Results

Preliminary case study-based applications of

FASTER-AI give initial evidence on its potential

impact, and the case studies show exactly how,

where, and what benefits FASTER-AI is likely to

work in practice. First, these cases apply to the very

nature of the different industries; hence applicability

and the effectiveness of the proposed framework will

be different in some characteristics. Specifically, the

electronic commerce site at stake is one of the world's

largest international retailers, with an extremely

varied client base, and it runs a state-of-the-art

recommendation engine. In the financial case study,

FASTER-AI was piloted on a long-established

institution that is highly regulatory inquisitive and

services a wide and diverse customer base. Finally,

the health-related case study was based on an

average-sized care provider heavily invested in AI-

supported clinical decision support systems. The final

assessed dimensions were the robustness and ethical

accountability for the AI-driven fraud detection

system of a financial institution.

3.2.1 Fairness in E-Commerce

Recommendation Systems

In an e-commerce platform's recommendation

system, the integration of fairness measures aimed at

improving the equitable distribution of

recommendations was tested. Over a three-month

period, the system was monitored using demographic

parity and disparate impact ratio metrics. As shown in

Table 1, the demographic parity improved by 20%,

suggesting a more balanced representation of

different user groups in the recommendations.

Additionally, customer feedback indicated a 15%

reduction in complaints related to perceived biases,

signalling an increase in user satisfaction with the

fairness of the system.

Table 1: Fairness Evaluation in E-Commerce

Recommendation System.

Metric Before

lementation

After

lementation

Chan

Demographic

arit

0.65 0.85 +20%

Customer

Complaints

(monthly)

120 102 -15%

3.2.2 Transparency in Financial AI Systems

Within finance, protocols for explainability made by

FASTER-AI were used on a credit scoring system

applied at a leading financial institution. This was

included for the combination of explainability tools

that would enable customers to fathom factors

affecting their credit score. Evaluation to include

customer satisfaction via surveys and disputes on loan

decisions. As would be seen in Table 2, these initial

results amount to a 15% reduction in disputes. This

also tended to coincide with improved levels of

customer satisfaction ratings. These would therefore

suggest that transparency initiatives increased user

confidence and contributed toward better regulatory

outcomes.

Table 2: Transparency Evaluation in Credit Scoring

System.

Metric Before

Implementation

After

Implementation

Change

Customer

Disputes

(

monthl

200

170

-15%

Customer

Satisfaction

Score

7.2/10 8.4/10 +1.2%

3.2.3 Security and Privacy in Healthcare AI

The security aspect of FASTER-AI was evaluated

within the context of healthcare, where artificial

intelligence models are used for clinical decision-

making support. The tests included state-of-the-art

encryption methods that are differential privacy

mechanisms for the protection of patient data. There

were no instances of leakage within the test period.

Therefore, this is the representation of how effective

the security methods adopted are. According to

medical professionals and patients through interviews

and surveying, increase in trust was realized when

sensitive protection information was catered to.

Therefore, the measures undertaken, though

resource-intensive as shown in Table 3, paid off-for

FASTER-AI: A Comprehensive Framework for Enhancing the Trustworthiness of Artiﬁcial Intelligence in Web Information Systems

389

there were no incidents of security, and stakeholders

gave favourable assessments.

Table 3: Security Evaluation in Healthcare AI System.

Metric Before

lementation

After

lementation

Chan

Data

Breaches

(

orted

)

2 0

100

Stakeholder

Confidence

Level

6.8/10 8.5/10 +1.7

3.2.4 Robustness in Fraud Detection

Systems

A controlled experimental framework is developed to

investigate the robustness of the AI model applied in

fraud detection. Adversarial training

recommendations, with respect to FASTER-AI, are

translated into practice under iterative changes,

considering multiple attack scenarios.

The results in Table 4 clearly indicated that the

model was indeed more resilient by 30%, with lower

error rates in an attack environment. These results

suggest FASTER-AI might be useful to improve

significantly the robustness of artificial intelligence

systems. Further testing in a variety of applications is

required to confirm these findings.

Table 4: Robustness Evaluation in Fraud Detection System.

Metric Before

Implementation

After

Implementation

Change

Error Rate

Under

Attack

12% 8%

-30%

Recovery

Time

(minutes)

15 10 -33%

3.2.5 Ethical Accountability in Financial

Institutions

Ethics accountability was done by developing an AI

ethics committee in a financial organization. This

committee's job was essentially to second-guess AI-

based decisions, especially the very sensitive ones

such as loan approvals. Early indications were that

input from the committee resulted in a more

consistent and explainable decision-making process.

As seen in Table 5, some of the ethical quagmires

were more easily avoided, and the organization had

far fewer external complaints about the

appropriateness of AI-led decisions. But the actual

effectiveness of those governance frameworks over

time will depend on their absorption into the greater

organizational culture.

Table 5: Ethical Accountability Evaluation in Financial

Institution.

Metric Before

Implementation

After

Implementation

Change

Ethical

Dilemma

Revolution

Time

5 days 3 days -40%

External

Complaints

(monthly)

15 9 -40%

Preliminary results from the deep-dive case

studies have given a few important pointers about

how effective FASTER-AI is for each of the aspects

of AI trustworthiness. In the case of an e-commerce

platform, improved demographic parity combined

with reduced customer complaints points to improved

fairness in AI-based recommendation systems

brought about by FASTER-AI. This implies that by

lowering the intrinsic biases, businesses could

increase consumer satisfaction while possibly

increasing participation by a wider consumer base.

The reduction in customer complaints after

transparency tools were implemented as part of the

credit scoring financial institutions mechanism

demonstrates well how FASTER-AI makes decisions

by artificial intelligence explainable and acceptable.

This has remarkable importance within the regulatory

framework, since often the requirements involve

transparency—the fact that FASTER-AI might help

financial institutions get closer to compliance with the

obligations while building customers' trust.

As the healthcare case study, the security

protocols recommended by FASTER-AI were able to

remain stable against data breaches and increase

stakeholder confidence. Even though implementation

required high resource involvement, the lack of

adverse security incidences together with positive

feedback from the stakeholders underpins the role

which such holistic security measures play in

sensitive industries such as healthcare. In this vein,

the system's frailty analyses evidenced very low error

rates after simulated attacks on fraud detection,

suggesting that FASTER-AI recommendations for

adversarial training of the poison or trigger models are

much more likely to result in stronger AI.

Overall, these results

should be particularly

encouraging in high-stakes settings where AI models

are often under attack by adversaries.

Finally, the setting-up of an AI ethics committee

within the bank supported these drivers because it

WEBIST 2024 - 20th International Conference on Web Information Systems and Technologies

390

allowed more standardized and transparent

implementation on ethical conflict resolution

practices. It showed thus the possibility of FASTER-AI to

institutionalize ethical responsibility. The decrease in

public grievances finally signals that governance

mechanisms along the lines should/can be crucial to build

confidence in AI-supported rule-deliberation.

4 CONCLUSIONS AND FUTURE

WORK

This paper proposes the FASTER-AI framework in

order to enhance AI trustworthiness in WIS along the

dimensions of fairness, transparency, security,

robustness, and ethical accountability. First case

studies conducted in various sectors have

demonstrated that the adoption of FASTER-AI

enhances the reliability of AI, since it provides higher

fairness, explainability, security, and ethical trust.

However, these initial findings relate to small

samples, and further research is expected to validate

the adaptation of FASTER-AI into larger and more

complex settings.

As AI evolves, so does FASTER-AI, bound to

proliferate with emerging challenges. Above all, an

effective collaboration between academic

institutions, industry players, and the regulators is

very instrumental in establishing common metrics

and standards for its evaluation. Long-term

implications of deploying FASTER-AI on

organizational change, user trust, and regulatory

compliance would, therefore, be an area of future

research, possibly through longitudinal studies.

Concluding, FASTER-AI contributes to the

debate on trustworthy AI by delivering a real-world

framework for WIS and hence laying the foundation

for creating and maintaining trust in AI systems; trust

will increasingly be necessary for efficacy and

societal acceptance of AI technologies.

REFERENCES

Agarwal, A., Beygelzimer, A., Dudík, M., Langford, J., &

Wallach, H. (2018). A reductions approach to fair

classification. International Conference on Machine

Learning, 60–69.

Bhutani, S., & Mittal, R. (2023). Blockchain Based

Ecommerce Warranty System Using NFT (Non-

fungible Token). 2023 3rd International Conference on

Advancement in Electronics & Communication

Engineering (AECE), 374–377.

Carlini, N., Liu, C., Erlingsson, Ú., Kos, J., & Song, D.

(2019). The secret sharer: Evaluating and testing

unintended memorization in neural networks. 28th

USENIX Security Symposium (USENIX Security 19),

267–284.

Danks, D., & London, A. J. (2017). Algorithmic Bias in

Autonomous Systems. Ijcai, 17(2017), 4691–4697.

Doshi-Velez, F., & Kim, B. (2017). Towards a rigorous

science of interpretable machine learning. ArXiv

Preprint ArXiv:1702.08608.

Floridi, L., Cowls, J., Beltrametti, M., Chatila, R.,

Chazerand, P., Dignum, V., Luetge, C., Madelin, R.,

Pagallo, U., & Rossi, F. (2018). AI4People—an ethical

framework for a good AI society: opportunities, risks,

principles, and recommendations. Minds and

Machines, 28, 689–707.

Floridi, L., Cowls, J., King, T. C., & Taddeo, M. (2020).

How to Design AI for Social Good: Seven Essential

Factors. Science and Engineering Ethics, 26(3), 1771–

1796. https://doi.org/10.1007/s11948-020-00213-5

Ganesh, D., & Rastogi, A. (2023). Integrating Web 3.0

Technologies for Advanced Information Retrieval

Systems in the Educational Sector. https://doi.org/

10.1109/ICCSAI59793.2023.10421149

Ge, M., Pilato, G., Persia, F., & D’Auria, D. (2023).

Recommender System for Social Media: Research

Challenges and Future Applications. https://doi.org/

10.1109/TransAI60598.2023.00033

Goodfellow, I., Shlens, J., & Szegedy, C. (2015).

Explaining and harnessing adversarial examples.

Huang, D. (2022). Design of Integrated English online

education system based on Artificial Intelligence.

https://doi.org/10.1109/ISoIRS57349.2022.00021

Kardaras, D. K., Troussas, C., Barbounaki, S. G., Tselenti,

P., & Armyras, K. (2024). A Fuzzy Synthetic

Evaluation Approach to Assess Usefulness of Tourism

Reviews by Considering Bias Identified in Sentiments

and Articulacy. Information, 15(4). https://doi.org/

10.3390/info15040236

Krouska, A., Troussas, C., & Sgouropoulou, C. (2020).

Applying Genetic Algorithms for Recommending

Adequate Competitors in Mobile Game-Based

Learning Environments. In V. Kumar & C. Troussas

(Eds.), Intelligent Tutoring Systems (pp. 196–204).

Springer International Publishing.

Madry, A., Makelov, A., Schmidt, L., Tsipras, D., & Vladu,

A. (2017). Towards Deep Learning Models Resistant to

Adversarial Attacks.

Martin, K. (2022). Ethics of data and analytics: Concepts

and cases. Auerbach Publications.

Mehrabi, N., Morstatter, F., Saxena, N., Lerman, K., &

Galstyan, A. (2021). A survey on bias and fairness in

machine learning. ACM Computing Surveys (CSUR),

54(6), 1–35.

Ribeiro, M. T., Singh, S., & Guestrin, C. (2016). “Why

should i trust you?” Explaining the predictions of any

classifier. Proceedings of the 22nd ACM SIGKDD

International Conference on Knowledge Discovery and

Data Mining, 1135–1144.

Rieke, N., Hancox, J., Li, W., Milletarì, F., Roth, H. R.,

Albarqouni, S., Bakas, S., Galtier, M. N., Landman, B.

A., Maier-Hein, K., Ourselin, S., Sheller, M., Summers,

FASTER-AI: A Comprehensive Framework for Enhancing the Trustworthiness of Artiﬁcial Intelligence in Web Information Systems

391

R. M., Trask, A., Xu, D., Baust, M., & Cardoso, M. J.

(2020). The future of digital health with federated

learning. Npj Digital Medicine, 3(1), 119. https://

doi.org/10.1038/s41746-020-00323-1

Ryan, M., & Stahl, B. C. (2020). Artificial intelligence

ethics guidelines for developers and users: clarifying

their content and normative implications. Journal of

Information, Communication and Ethics in Society,

19(1), 61–86.

Shahriari, K., & Shahriari, M. (2017). IEEE standard

review—Ethically aligned design: A vision for

prioritizing human wellbeing with artificial intelligence

and autonomous systems. 2017 IEEE Canada

International Humanitarian Technology Conference

(IHTC), 197–201.

Troussas, C., Krouska, A., & Sgouropoulou, C. (2022).

Enriching mobile learning software with interactive

activities and motivational feedback for advancing

users’ high-level cognitive skills. Computers, 11(2), 18.

Troussas, C., Virvou, M., & Espinosa, K. J. (2015). Using

visualization algorithms for discovering patterns in

groups of users for tutoring multiple languages through

social networking. J. Networks, 10(12), 668–674.

Veale, M., & Binns, R. (2017). Fairer machine learning in

the real world: Mitigating discrimination without

collecting sensitive data. Big Data & Society, 4(2),

2053951717743530.

Virvou, M., Troussas, C., Caro, J., & Espinosa, K. J. (2012).

User Modeling for Language Learning in Facebook. In

P. Sojka, A. Horák, I. Kopeček, & K. Pala (Eds.), Text,

Speech and Dialogue (pp. 345–352). Springer Berlin

Heidelberg.

Zenkert, J., & Fathi, M. (2023). Practice-Oriented

Approaches for Information and Metadata Management

in a Content Management System-Learnings from the

Smart City Project LOKAL-digital. 2023 International

Conference on Electrical, Computer and Energy

Technologies (ICECET), 1–6.

WEBIST 2024 - 20th International Conference on Web Information Systems and Technologies

392