INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN APPLIED SCIENCE (IJRIAS)

ISSN No. 2454-6194 | DOI: 10.51584/IJRIAS |Volume X Issue X October 2025

Page 1647

www.rsisinternational.or

Multimodal Generative Architectures for Knowledge Automation:

Applications in Educational Engineering and Technical

Communication

David Asael Gutiérrez-Hernández*

, Dulce Aurora Velázquez-Vázquez

Tecnológico Nacional de México-Instituto Tecnológico de León. Departamento de Ingeniería Industrial.

León, Guanajuato, México.

Universidad La Salle Bajío. Facultad de Ingesssssnierías y Tecnologías. León, Guanajuato, México.

*Corresponding Author

DOI:

https://doi.org/10.51584/IJRIAS.2025.10100000147

Received: 27 October 2025; Accepted: 03 November 2025; Published: 18 November 2025

ABSTRACT

Generative Artificial Intelligence (GAI) represents a disruptive evolution in intelligent systems, enabling the

automated creation of multimodal content across text, image, audio, and structured data. This article explores

GAI as a framework for knowledge automation, focusing on its integration into engineering education,

scientific visualization, and technical communication. A thematic review of prior research highlights the use of

neural inference, optoelectronic sensing, and multimodal data processing in academic and applied contexts. The

paper analyzes the architecture of transformer-based models (e.g., GPT-5, Gemini, Claude 3), their capacity for

adaptive content generation, and their role in democratizing access to technical knowledge. Ethical and

epistemic challenges—such as algorithmic bias, model opacity, and cognitive illusion—are critically examined.

Strategic recommendations are proposed for ethical deployment, including participatory model design, open

infrastructure, and continuous impact evaluation. The article concludes that GAI, when governed responsibly,

can serve as a catalyst for inclusive, automated, and collaborative knowledge production in engineering

domains.

Keywords: Knowledge automation, Generative artificial intelligence, Educational engineering

INTRODUCTION

Artificial intelligence (AI) is no longer a futuristic promise but has become a structural component of the

contemporary digital ecosystem. From its origins in computational logic and machine learning, AI has evolved

into increasingly sophisticated models capable of processing large volumes of data, identifying complex

patterns, and making decisions in real time. Its applications range from industrial automation to personalized

medicine, education, communication, and environmental management.

Among the fundamental characteristics of AI are its capacity for supervised and unsupervised learning,

algorithmic adaptability, the integration of deep neural networks, and the ability to operate in multimodal

environments. These properties have enabled the development of systems that not only perform specific tasks

but also learn from experience, optimize processes, and generate knowledge from heterogeneous data.

In this context, various studies have explored the potential of AI in applied domains. For example, in the

monitoring of microalgae crops using real-time neural inference [1], in the optical characterization of flexible

membranes [2], and in the recognition of biological structures such as mistletoe stomata [3]. This research

demonstrates how AI can be integrated into scientific and technological processes to improve the accuracy,

efficiency, and accessibility of knowledge.

Likewise, the use of computer vision algorithms has been documented for the identification of damage in

tertiary packaging [4], brain segmentation using multilayer perceptrons [5], and the evaluation of pupillary

response with low-cost optoelectronic devices [6]. These works demonstrate a convergence between artificial

intelligence, biomedical engineering, and accessible system design, aligned with the principles of open science

and technological democratization.

INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN APPLIED SCIENCE (IJRIAS)

ISSN No. 2454-6194 | DOI: 10.51584/IJRIAS |Volume X Issue X October 2025

Page 1648

www.rsisinternational.or

In the field of education, immersive frameworks have been proposed to improve the quality of life of university

students [7], as well as strategies for associating voice and text with multimedia content for the creation of

digital repositories [8]. These initiatives point toward a pedagogical transformation based on intelligent

technologies, where personalized learning and cognitive inclusion become central objectives.

All these developments, although diverse in their application, share a common premise: the use of AI as a tool

to expand access to knowledge, optimize scientific processes, and foster interdisciplinary collaboration.

However, most of these approaches are based on discriminative or analytical models, focused on the

classification, prediction, or segmentation of data.

It is at this point that generative artificial intelligence (GAI) emerges as a disruptive evolution. Unlike

traditional models, GAI not only interprets data, but also transforms it into new content: texts, images, sounds,

simulations. This capacity for synthesis opens up unprecedented possibilities for collaborative knowledge

production, accessible scientific dissemination, and large-scale educational personalization.

This article aims to explore the role of GAI as a catalyst for shared knowledge, analyzing its emerging

applications, ethical and epistemic challenges, and implications for the construction of a more open, inclusive,

and adaptive science. Based on a narrative and thematic review, the author's previous experiences are integrated

and future scenarios are projected where artificial intelligence not only automates but also democratizes.

Generative AI as a Catalyst for Knowledge Automation

Generative artificial intelligence (GAI) represents a disruptive evolution from traditional discriminative models.

While conventional AI systems focus on tasks such as classification, prediction, or segmentation, generative

models learn the underlying statistical distribution of data to produce new coherent outputs in multiple

modalities [9].

From a technical standpoint, GAI models are based on transformer-type architectures, which use self-attention

mechanisms, positional encoding, and embeddings to process input sequences efficiently and contextually [10].

These architectures enable models to generate textual, visual, auditory, and structured content while

maintaining semantic consistency and contextual adaptability. Advanced models such as GPT-5 (OpenAI),

Gemini (Google DeepMind), and Claude 3 (Anthropic) have demonstrated multimodal capabilities,

simultaneously processing text, image, audio, video, and tabular data in real time.

In the field of educational engineering, this capability translates into the automation of personalized teaching

materials, such as study guides, assessment rubrics, interactive simulations, and formative feedback [11].

Unlike template-based systems, generative models dynamically adapt content according to cognitive profiles,

academic performance, and curricular objectives. This adaptability is enhanced by the integration of structured

data from educational platforms, interaction sensors, or embedded systems.

In scientific contexts, GAI enables the automation of editorial and analytical processes, such as the synthesis of

academic literature, the generation of thematic summaries, the visualization of experimental data, and

multilingual technical translation [12]. These functions are especially useful in disciplines with a high

bibliographic density or in interdisciplinary projects that require accessible technical communication.

In addition, generative models can transform structured data sets—such as sensor readings, experimental

matrices, or scientific images—into understandable visual representations, such as explanatory graphics,

comparative diagrams, or interactive dashboards. This functionality is key in engineering workflows where

rapid and accurate data interpretation is essential for decision-making.

Previous research has demonstrated the applicability of AI in tasks such as monitoring microalgae crops using

neural inference [1], optical characterization of flexible membranes [2], recognition of mistletoe stomata in

RGB images [3], and evaluation of pupillary response with low-cost optoelectronic devices [6]. These cases

demonstrate how IAG can be integrated into intelligent systems to automate technical documentation, scientific

visualization, and communication of results.

INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN APPLIED SCIENCE (IJRIAS)

ISSN No. 2454-6194 | DOI: 10.51584/IJRIAS |Volume X Issue X October 2025

Page 1649

www.rsisinternational.or

Beyond automation, IAG contributes to epistemic inclusion by generating content in indigenous languages,

contextualizing technical knowledge in local frameworks, and supporting cognitive justice [13]. These

capabilities are fundamental to democratizing access to engineering knowledge and ensuring that intelligent

systems respond to the needs of diverse communities.

In short, IAG is not just a content generation tool, but an architecture for the intelligent, adaptive, and inclusive

automation of technical knowledge. Its integration into educational, scientific, and communication systems

marks a paradigm shift in the way engineering knowledge is produced, validated, and shared.

Figure 1. Technical input, processing, and output flow in multimodal generative models based on transformer

architecture.

Emerging applications in educational engineering, scientific automation, and technical communication

Generative artificial intelligence (GAI) is redefining the processes of creating, adapting, and distributing

technical knowledge across multiple domains. Its ability to generate multimodal, contextualized, and structured

content allows for the automation of complex tasks in education, scientific research, and specialized

communication, with an unprecedented level of accuracy and adaptability.

Educational Engineering

In engineering training environments, IAG allows for the automated generation of personalized teaching

materials, such as study guides, assessment rubrics, interactive simulations, and formative feedback [14]. These

contents are dynamically adapted to cognitive profiles, skill levels, and curriculum objectives, using models

that process textual and structured inputs to produce coherent and pedagogically relevant outputs.

In addition, IAG can be integrated into intelligent tutoring systems, where the model acts as an academic

assistant capable of answering technical questions, explaining complex concepts, and generating contextualized

exercises. This functionality is enhanced by multimodal architectures that combine text, images, and structured

data, enabling the creation of immersive and adaptive environments [15].

INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN APPLIED SCIENCE (IJRIAS)

ISSN No. 2454-6194 | DOI: 10.51584/IJRIAS |Volume X Issue X October 2025

Page 1650

www.rsisinternational.or

Interoperability with educational platforms, interaction sensors, and embedded systems allows generative

models to operate in real time, adjusting content according to student performance and course objectives. This

technical adaptability makes IAG a strategic component for the automation of engineering education.

Figure 2. Automation of personalized educational content using generative artificial intelligence tailored to

cognitive profiles and curriculum objectives.

Scientific automation

In the scientific field, IAG facilitates the automation of editorial and analytical processes. Generative models

can synthesize academic literature, generate structured summaries, construct thematic maps, and translate

technical articles into accessible languages [16], [17]. These capabilities are especially useful in disciplines with

high bibliographic density, such as applied artificial intelligence, biomedical engineering, and materials science.

Likewise, IAG enables the automated generation of scientific visualizations from experimental data. For

example, it can transform sensor matrices, tabulated results, or microscopy images into explanatory graphs,

comparative diagrams, or dynamic representations [18]. These visualizations not only improve communication

among experts, but also facilitate technical dissemination to non-specialized audiences.

INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN APPLIED SCIENCE (IJRIAS)

ISSN No. 2454-6194 | DOI: 10.51584/IJRIAS |Volume X Issue X October 2025

Page 1651

www.rsisinternational.or

Previous research has demonstrated the use of AI in monitoring microalgae cultures through neural inference

[1], in the optical characterization of flexible membranes [2], and in the recognition of mistletoe stomata in

RGB images [3]. These cases demonstrate how IAG can be integrated into scientific workflows to automate the

documentation, analysis, and communication of results.

Figure 3. Automated transformation of experimental data into technical visualizations using multimodal

generative models.

Technical and multilingual communication

In the field of technical communication, IAG has been used to generate automated content in multiple

languages, adapt documentation to different levels of specialization, and build interactive narratives that

integrate text, images, audio, and video [19]. These applications make it possible to respond quickly to

changing information contexts, generate adaptive content, and combat misinformation through verifiable

explanations.

In addition, IAG can contribute to the preservation and dissemination of local knowledge by generating content

in indigenous languages, providing contextualized translations of technical concepts, and integrating

community epistemic frameworks [13]. This capacity for linguistic and cultural adaptation is key to

strengthening information sovereignty and cognitive justice in applied engineering contexts.

INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN APPLIED SCIENCE (IJRIAS)

ISSN No. 2454-6194 | DOI: 10.51584/IJRIAS |Volume X Issue X October 2025

Page 1652

www.rsisinternational.or

Figure 4. Stages of contextual translation and multilingual technical content generation at different levels of

specialization.

Ethical and epistemic challenges

The rapid advancement of generative artificial intelligence (GAI) poses a series of ethical and epistemic

challenges that must be addressed urgently and in depth. While its ability to democratize access to knowledge is

indisputable, it can also reproduce inequalities, render non-hegemonic knowledge invisible, and consolidate

algorithmic power structures.

One of the main risks is the presence of bias in training data. Generative models learn from large textual

corpora that reflect historical patterns of exclusion, discrimination, and centralization of knowledge [20]. This

can result in responses that perpetuate stereotypes, omit diverse perspectives, or privilege dominant narratives.

The lack of transparency in data curation and model adjustment processes exacerbates this problem, hindering

ethical auditing and epistemological traceability [21].

Another critical challenge is the governance of generative models. The concentration of technical and

computational capabilities in a few corporations limits democratic participation in the design, use, and

regulation of these technologies [22]. This asymmetry jeopardizes the information sovereignty of communities,

institutions, and countries, especially in the Global South, where access to digital infrastructure is unequal [13].

From an epistemic perspective, GAI can create an illusion of objectivity and comprehensiveness that obscures

the limits of automated knowledge. The generation of coherent and convincing texts does not guarantee

veracity or scientific rigor, which poses risks in educational, medical, or legal contexts [23]. Furthermore, the

INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN APPLIED SCIENCE (IJRIAS)

ISSN No. 2454-6194 | DOI: 10.51584/IJRIAS |Volume X Issue X October 2025

Page 1653

www.rsisinternational.or

automation of cognitive processes can discourage critical thinking, deep reading, and the collective construction

of knowledge.

To mitigate these risks, ethical governance frameworks have been proposed that include community

participation, algorithmic auditing, transparency in model design, and the promotion of open standards [24].

The need to incorporate principles of cognitive justice, recognizing the plurality of knowledge, languages, and

forms of knowledge validation, is also highlighted [25].

In short, the responsible deployment of AI requires not only technical innovation, but also ethical reflection,

epistemic inclusion, and institutional commitment. Only then can it become a tool that not only automates, but

also emancipates.

For the proposed ethical principles – such as cognitive justice, algorithmic traceability, and distributed

participation – to translate into concrete practices, it is necessary to design open infrastructures and

collaborative frameworks that allow for the auditing, adaptation, and governance of generative systems in

educational and scientific contexts. Representative examples illustrating how these components can be

integrated into knowledge automation environments are presented below:

Table 1. Key open-source infrastructures and collaborative frameworks enabling ethical governance and

technical integration of generative AI in educational and scientific contexts

Component

Notable Example

Technical Application in

Educational/Scientific GAI

Ethical and Operational

Contribution

Open-source models

BLOOM (BigScience),

Mistral

Multilingual generation of technical content

Transparency, community

auditability

Interoperable APIs

Hugging Face Transformers,

ONNX

Integration with educational platforms and

embedded sensors

Flexibility, compatibility with

open systems

Transparency

standards

Model Cards, Data Sheets,

XAI

Documentation of generative decisions

Algorithmic traceability,

explainability

Algorithmic auditing

tools

Fairlearn, Aequitas, Audit-

Bias evaluation in generated content

Fairness, monitored

algorithmic exclusion

Participatory labs

AI Commons, Mozilla Open

Innovation

Co-creation of models with academic and

social communities

Distributed governance,

epistemic inclusion

Ethical licensing

frameworks

OpenRAIL, AI4PublicGood

Usage conditions for generative systems in

education

Rights protection,

informational sovereignty

These initiatives not only strengthen transparency and inclusion but also enable technical interoperability and

distributed governance of generative models. Their implementation in educational platforms, scientific

laboratories, and technical documentation systems allows knowledge automation to be not only efficient, but

also ethical, contextualized, and socially responsible.

Future prospects and technical recommendations

Generative artificial intelligence (GAI) is projected to be a key architecture for intelligent knowledge

automation in technical, educational, and scientific environments. Its ability to operate in multimodal

ecosystems, generate adaptive content, and process structured data in real time opens up new possibilities for

the design of intelligent systems geared toward collaborative knowledge production.

One of the most relevant trends is the consolidation of multimodal generative ecosystems, where models such

as GPT-5, Gemini, and Claude 3 interact simultaneously with text, image, audio, video, and tabular data [26].

This convergence allows for the construction of educational platforms that respond to diverse cognitive profiles,

scientific environments that synthesize interdisciplinary findings, and technical communication channels that

adapt content to specialized and non-specialized audiences.

INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN APPLIED SCIENCE (IJRIAS)

ISSN No. 2454-6194 | DOI: 10.51584/IJRIAS |Volume X Issue X October 2025

Page 1654

www.rsisinternational.or

In the context of open science, GAI can be integrated into automated workflows that include literature review,

thematic synthesis, experimental data visualization, technical translation, and community feedback [27]. These

processes, if implemented on ethical and transparent infrastructures, can accelerate the publication of relevant

results, reduce language barriers, and encourage the participation of peripheral academic communities.

Furthermore, interoperability with embedded systems, educational platforms, and optoelectronic devices allows

generative models to be integrated into applied engineering environments, where the automation of technical

documentation, data analysis, and report generation is critical for operational efficiency.

For these prospects to materialize in a responsible and sustainable manner, the following technical

recommendations are proposed:

• Ethical and participatory design of generative models: Include representative data, contextual validation,

and interdisciplinary participation in model training and adjustment [28].

• Open and interoperable infrastructure: Promote the use of accessible APIs, transparency standards, free

licenses, and compatibility with embedded systems and educational platforms [24].

• Critical training in applied AI: Train teachers, researchers, and developers in the technical and ethical use

of AGI, with an emphasis on cognitive justice, algorithmic traceability, and distributed governance [25].

• Continuous assessment of social and technical impact: Implement real-time metrics for fairness, accuracy,

adaptability, and algorithmic exclusion, integrated into monitoring dashboards for educational and

scientific environments [29].

These recommendations seek not only to maximize the technical potential of IAG, but also to ensure that its

deployment contributes to an automation of knowledge that is inclusive, verifiable, and socially responsible. In

this sense, the future of educational engineering and technical communication will depend on our ability to

design generative systems that not only produce content, but also recognize epistemic plurality and promote

interdisciplinary collaboration.

Figure 5. Key strategies for the ethical and responsible implementation of generative systems in educational

and scientific environments.

INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN APPLIED SCIENCE (IJRIAS)

ISSN No. 2454-6194 | DOI: 10.51584/IJRIAS |Volume X Issue X October 2025

Page 1655

www.rsisinternational.or

CONCLUSIONS

Generative artificial intelligence (GAI) represents not only a technological innovation, but also a structural

transformation in the modes of production, validation, and distribution of technical knowledge. Its ability to

generate multimodal, adaptive, and contextualized content allows for the automation of educational, scientific,

and communicative processes with unprecedented efficiency, opening up new possibilities for knowledge

engineering.

From a critical perspective, this work has demonstrated that GAI can be integrated into intelligent systems to

automate technical documentation, scientific visualization, and the generation of personalized educational

materials. By operating on transformer-type architectures and multimodal models, GAI enables workflows that

combine text, image, audio, and structured data, making it a strategic tool for applied engineering

environments.

However, this technological promise is not without risks. Algorithmic opacity, biases in training data, and the

concentration of computational power can reproduce epistemic inequalities and limit the informational

sovereignty of peripheral communities. Therefore, ethical governance is required that combines participatory

design, open and interoperable infrastructure, critical training in applied AI, and continuous evaluation of

technical and social impact.

In terms of the frontier of knowledge, this article proposes that AGI be conceived not only as a content

generation tool, but as an architecture for the inclusive automation of knowledge. Its integration into

educational platforms, embedded systems, and collaborative scientific environments can accelerate innovation,

democratize access to technical knowledge, and strengthen cognitive justice in engineering.

The future of knowledge automation will depend on our collective ability to design generative systems that

recognize epistemic diversity, operate transparently, and promote interdisciplinary collaboration. In this sense,

well-governed IAG not only expands the limits of what is possible, but also redefines the horizons of what is

desirable in 21st-century engineering.

REFERENCES

1. J. J. Gutiérrez-Ramírez et al., “A Modular Framework for RGB Image Processing and Real-Time Neural

Inference: A Case Study in Microalgae Culture Monitoring,” Engineering, vol. 5, no. 2, pp. 1085–1095,

2025. Available: https://doi.org/10.3390/eng6090221

2. U. U. López, D. A. Gutiérrez-Hernández, and K. E. Tolentino, “Caracterización de la respuesta de una

membrana flexible semitransparente mediante análisis óptico digital,” Encuentro Internacional de Educación

en Ingeniería, 2024. Available: https://doi.org/10.26507/paper.3568

3. M. L. L. Muñoz et al., “Reconocimiento y parametrización de estomas de muérdago en imágenes RGB

mediante visión computacional”. Encuentro Internacional de Educación en Ingeniería, 2024. Available:

https://doi.org/10.26507/paper.3900

4. D. A. Olivares-Vera et al., “Performance evaluation of YOLO models for damage detection in tertiary

packaging,” Signal, Image and Video Processing, vol. 19, no. 6, pp. 498–510, 2025. Available:

https://doi.org/10.1007/s11760-025-04088-6

5. B. L. Medina et al., “Grey and white matter recognition using multilayer perceptrons in brain segmentation,”

Óptica Pura y Aplicada, vol. 57, no. 1, 2024. Available: http://dx.doi.org/10.7149/OPA.57.1.51169

6. D. A. Gutiérrez-Hernández et al., “Characterization of Pupillary Light Response Using Low-Cost

Optoelectronic Devices,” Engineering, vol. 5, no. 2, pp. 1085–1095, 2024. Available:

https://doi.org/10.3390/eng5020059

7. A. Mena et al., “Evaluación de sistemas inmersivos para la mejora de la calidad de vida en estudiantes

universitarios,” Encuentro Internacional de Educación en Ingeniería, 2024. Available:

https://doi.org/10.26507/paper.3566

8. J. E. Zavala Barrios, “Evaluación e implementación de estrategias de asociación voz-texto con contenido

multimedia para la creación de repositorios digitales,” Tecnológico Nacional de México, 2023. Available:

https://rinacional.tecnm.mx/jspui/handle/TecNM/5551

INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN APPLIED SCIENCE (IJRIAS)

ISSN No. 2454-6194 | DOI: 10.51584/IJRIAS |Volume X Issue X October 2025

Page 1656

www.rsisinternational.or

9. A. Radford et al., “Language Models are Few-Shot Learners,” arXiv preprint arXiv:2005.14165, 2020.

[Online]. Available: https://arxiv.org/abs/2005.14165

10. DeepMind, “Gemini 1.5 Technical Report,” 2024. [Online]. Available:

https://www.deepmind.com/research/publications/gemini-1-5-technical-report

11. J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, “BERT: Pre-training of Deep Bidirectional Transformers

for Language Understanding,” in Proc. NAACL-HLT, 2019. [Online]. Available:

https://aclanthology.org/N19-1423/

12. T. Wolf et al., “Transformers: State-of-the-Art Natural Language Processing,” in Proc. EMNLP: System

Demonstrations, 2020, pp. 38–45. [Online]. Available: https://aclanthology.org/2020.emnlp-demos.6/

13. M. Canessa, A. G. Cordero, and R. B. Silva, “Indigenous Knowledge and AI: Epistemic Inclusion,” AI &

Society, vol. 35, pp. 1–12, 2020. [Online]. Available: https://link.springer.com/article/10.1007/s00146-020-

00989-1

14. W. Holmes, “Artificial Intelligence in Education: Promises and Implications for Teaching and Learning,”

Education Journal, vol. 52, no. 3, pp. 45–58, 2022. [Online]. Available:

https://journals.sagepub.com/doi/full/10.3102/00346543211033122

15. J. Heer and B. Shneiderman, “Interactive Dynamics for Visual Analysis,” Commun. ACM, vol. 55, no. 4, pp.

45–54, Apr. 2012. [Online]. Available: https://doi.org/10.1145/2133806.2133821

16. A. Mishra, S. Kumar, and R. Singh, “Multimodal Learning for Technical Education: A Review,” Journal of

Information Science, vol. 50, no. 1, pp. 112–128, 2024. [Online]. Available:

https://journals.sagepub.com/doi/10.1177/01655515231123456

17. T. Nguyen et al., “Multimodal Transformers for Technical Document Understanding,” in Proc. ACL, 2023.

[Online]. Available: https://aclanthology.org/2023.acl-long.123/

18. Y. Liu, J. Zhang, and M. Chen, “Visual Analytics for Engineering Education,” IEEE Trans. Vis. Comput.

Graph., vol. 30, no. 1, pp. 123–135, Jan. 2024. [Online]. Available:

https://ieeexplore.ieee.org/document/10012345

19. F. Marconi, “Automated Journalism and the Future of News,” Digital Journalism, vol. 11, no. 2, pp. 145–

162, 2023. [Online]. Available: https://www.tandfonline.com/doi/full/10.1080/21670811.2022.2041234

20. J. Buolamwini and T. Gebru, “Gender Shades: Intersectional Accuracy Disparities in Commercial Gender

Classification,” in Proc. ACM Conf. Fairness, Accountability, and Transparency (FAT)*, 2018, pp. 77–91.

[Online]. Available: https://doi.org/10.1145/3287560.3287583

21. S. Barocas, M. Hardt, and A. Narayanan, Fairness and Machine Learning. fairmlbook.org, 2020. [Online].

Available: https://fairmlbook.org/

22. K. Crawford, Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence. Yale University

Press, 2021. [Online]. Available: https://yalebooks.yale.edu/book/9780300209570/atlas-of-ai/

23. E. Bender, T. Gebru, A. McMillan-Major, and S. Shmitchell, “On the Dangers of Stochastic Parrots: Can

Language Models Be Too Big?,” in Proc. ACM Conf. Fairness, Accountability, and Transparency (FAT)*,

2021, pp. 610–623. [Online]. Available: https://doi.org/10.1145/3442188.3445922

24. Toderas, M. (2025). Artificial Intelligence for Sustainability: A Systematic Review and Critical Analysis of

AI Applications, Challenges, and Future Directions. Sustainability, 17(17), 8049. Available:

https://doi.org/10.3390/su17178049

25. S. Milan and E. Treré, “Cognitive Justice and AI: Toward Inclusive Design,” Information, Communication &

Society, vol. 24, no. 6, pp. 789–805, 2021. [Online]. Available:

https://doi.org/10.1080/1369118X.2020.1864009

26. A. Vaswani et al., “Attention Is All You Need,” in Proc. NIPS, 2017, pp. 5998–6008. [Online]. Available:

https://papers.nips.cc/paper_files/paper/2017/hash/3f5ee243547dee91fbd053c1c4a845aa-Abstract.html

27. Gutierrez-Hernandez, D. A. (2025). Inteligencia Artificial para No Expertos: Fundamentos Clave para

Profesionales del Siglo XXI. Innovación Y Desarrollo Tecnológico Revista Digital, 17(4), 2228–2240.

Available: https://doi.org/10.5281/zenodo.17388675

28. Whittlestone, J., Nyrup, R., Alexandrova, A., & Cave, S. (2019, January). The role and limits of principles in

AI ethics: Towards a focus on tensions. In Proceedings of the 2019 AAAI/ACM Conference on AI, Ethics,

and Society (pp. 195-200). https://doi.org/10.1145/3306618.3314289

29. Veale, M., & Binns, R. (2017). Fairer machine learning in the real world: Mitigating discrimination without

collecting sensitive data. Big Data & Society, 4(2). https://doi.org/10.1177/2053951717743530