PREreview of Artificial Intelligence and Sustainability – A Review

Artificial Intelligence and Sustainability – A Review

Key words: AI, Sustainability, Research

This document is a review on artificial intelligence and sustainability, highlighting the economic, social and environmental perspectives, as well as the challenges and opportunities related to the use of AI for sustainable development.

Introductory content

This article, published on December 27, 2023, examines the impact of artificial intelligence on sustainability. It is available in open access under the Creative Commons Attribution license.

1. Introduction

This section examines the impact of AI on sustainability by analyzing economic, social and environmental dimensions.

This section discusses the advances of AI and its potential to transform various sectors such as healthcare, transportation, agriculture, energy and media.

It highlights the importance of examining the impact of AI on sustainability, taking into account economic, social and environmental dimensions.

The main objective is to create a systematic mapping study to compile and analyze the available literature on AI and sustainability, identifying knowledge gaps

2. Initial Literature Review

The section explores the impact of AI on sustainability, the challenges and opportunities, and the future research needed.

AI is defined as the ability of computer systems to perform tasks normally requiring human intelligence, with definitions provided by the Oxford dictionary and John McCarthy.

Sustainability is defined by the United Nations Brundtland Commission as meeting the needs of the present without compromising those of future generations, with three pillars: social, economic and environmental.

The concept of “Sustainable AI” addresses the use of AI to achieve sustainability and the sustainability of AI systems themselves.

Research shows that AI can reduce the ecological footprint but faces challenges such as dependence on historical data, uncertain human behavioral responses, and cybersecurity risks.

Future studies should include multilevel approaches, system dynamics, and psychological, sociological and economic considerations.

Applications of AI for sustainability cover diverse sectors such as construction, transportation, healthcare, manufacturing, agriculture and water management.

The “Green AI” concept focuses on the sustainable nature of AI, with research into compact networks, energy-efficient training strategies, and the efficient use of data.

3. Research Methodology

This section explores the sustainability of AI through a systematic mapping study, analyzing environmental, economic and social impacts.

This study uses a systematic mapping study (SMS) to explore the sustainability of AI, focusing on environmental, economic and social impacts.

The methodology follows the “input-processing-output” approach of Levy and J. Ellis, and includes preliminary research into the sustainability of AI.

Research questions include capturing AI sustainability in existing literature, the maturity level of the research field, and the future research agenda.

An exhaustive literature search was conducted using databases such as Google Scholar, Science Direct, IEEE, and ACM Digital Library.

Explicit inclusion and exclusion criteria were applied, resulting in the selection of 88 articles out of 148 initial candidates.

The results show that the majority of research focuses on AI's contributions to sustainability goals or on AI's impacts in specific fields.

Research categories were defined according to publication frequencies, identifying areas requiring future research.

4. Results and Visualization

The section explores the sustainability of AI, its environmental, social and economic impacts, and the evolution of research since 2019.

The section examines AI sustainability in two main categories: “AI sustainability” and “AI for sustainability”, each accounting for 43.2% of the 88 articles studied.

A third, more holistic category combines both aspects and includes 12 articles published mainly since 2019.

The sustainability dimensions addressed are environmental (26 articles), social (21 articles) and economic (5 articles), with a majority of articles integrating all three dimensions.

AI sustainability research saw a significant increase from 2019 onwards, with a peak in 2020 and 2021.

The types of article contributions mainly include philosophical research (33%), evaluation research (26%) and solution proposals (24%).

Analysis of the maturity of the field shows an increasing diversity of authors and research methods, with 19% of articles being empirical.

5.Discussion and Limitations

The section 5 highlights the importance of an integrated, multidimensional approach to assessing AI sustainability, using the Sustainable Development Goals (SDGs) as a framework.

Future research should continue to increase in number and diversity, with an emphasis on empirical analysis and author diversity to achieve a higher maturity.

The section discusses validity limitations, paper quality, strict analysis, qualitative evaluation and subjectivity.

The section addresses construct and conclusion validity limitations, as well as the quality and depth of selected papers.

Strict and focused analysis, while rigorous, may limit the generalizability of results to the whole field of AI sustainability.

Qualitative assessment and subjectivity in the authors' analysis may introduce bias and affect the accuracy of conclusions.

Despite these limitations, the research contributes to a better understanding of AI sustainability and identifies areas for future exploration.

Conclusion

The study analyzes the sustainability of AI by examining economic, social and environmental dimensions, and the evolution of research.

The study aims to create a systematic mapping study to analyze the sustainability of AI by focusing on economic, social and environmental dimensions.

AI sustainability research is divided between “AI as a tool for sustainability” and “AI's impact on sustainability”.

Prior to 2019, the field was relatively immature with few publications, but after this date the number of publications increased, indicating a maturing of the field.

The diversity of authors and the increase in empirical articles also contribute to the growing maturity of the field.

Competing interests

The author declares that they have no competing interests.