Artificial intelligence (AI) has become an essential component of many human activities, transforming sectors as various as health, finance, and marketing. However, its rapid evolution has raised fundamental questions regarding its impact. How can artificial intelligence be reconciled with the environment and ethics?
Biases in data can lead to unfair decisions, particularly in the areas of recruiting or finance. From the impact of automated results to the carbon footprint generated by AI models, responsible adoption is indispensable to guaranteeing that this technology benefits everyone without compromising either the equilibrium of our environment or our social values.
Tools such as marketing mix modelling (MMM) and generative AI models illustrate the potential of AI when it is exploited in a sustainable manner. This article explores the key issues related to AI and offers strategies for balancing technological innovation with ethics and sustainability.
Evaluating the social and environmental impact of artificial intelligence
Ensuring transparency and inclusivity
Transparency in AI systems is crucial for identifying and correcting biases or errors before they lead to prejudiced outcomes. Companies must also consider the obsolescence of certain occupations in light of automation and guide their teams by offering adapted training.
AI tools must be accessible to everyone, including people with disabilities and those living in regions with weaker Internet connectivity. A respect for ethical standards, particularly in terms of data protection and non-discrimination, is indispensable. Collaborating with experts in sociology, law, and the environment can also minimize negative impacts.
Lastly, informing users on the features and limitations of AI systems will strengthen their acceptance by society.
Analyzing and reducing the environmental impact
Data centres are major energy consumers and contribute significantly to carbon emissions. However, using renewable energy sources and reducing data movement over networks can limit unwanted effects while improving data security.
Companies need to track the environmental impact of their models using adapted metrics and change their strategies based on the latest technological advancements. Adopting recognized standards frameworks, such as those for the Association française de normalisation (AFNOR), can also help with AI initiatives.
Establishing responsible governance of artificial intelligence
In a world in which artificial intelligence is playing an increasingly central role, it is becoming essential to establish responsible governance for overseeing AI development and use. Such initiatives are based on three key pillars: the creation of clear internal policies, the training of teams, and the regular evaluation of the effects of AI solutions.
Creating clear internal policies for managing the use of AI
The first step towards effective governance involves defining explicit internal policies. These policies must specify limits and expectations in terms of AI use. For example, they must include ethical principles such as:
- transparency;
- fairness in data processing; and
- respect for privacy.
By establishing well-defined rules, organizations can prevent misuse and ensure that AI is aligned with their values and strategic goals.
Training teams in responsible, ethical AI practices
Responsibility in the use of AI also involves education and sensitization. Employees at all levels must be trained in the ethics issues and best practices associated with AI. This covers topics that include:
- avoiding algorithmic bias;
- protecting sensitive data; and
- decision-making for minimal impact on vulnerable populations.
Continual training can strengthen a team’s skills and help members navigate the constantly evolving technological landscape.
Regular evaluation of the impact of AI solutions
Lastly, responsible governance involves periodic evaluation of the impacts of artificial intelligence. This enables verification to determine whether the implemented solutions respect the organization’s social and environmental responsibility goals. These evaluations must include key indicators such as the environmental impact of AI systems, their contribution to creating social equity, and their compliance with current legislation. By integrating these reviews into their processes, organizations can quickly adapt in order to correct any deviation from these standards.
Adopting responsible governance for AI is not just an ethical requirement, it’s also an opportunity to create lasting value. This approach strengthens trust on the part of stakeholders and maximizes the benefits of AI while minimizing the risks associated with this revolutionary technology.
Choosing responsible models of generative AI
Evaluating the energy efficiency of models
Before deploying a model, it’s important to evaluate its energy consumption with the help of benchmarks. The use of pre-trained or shared models on responsible cloud infrastructure reduces the amount of resources needed. In simple situations, optimized, lightweight models that require fewer calculations to generate results are often sufficient.
Favouring modularity and model reuse
Opting for modular or reusable models limits the need to train separate systems for every task. For example, using base models that are sector-specific reduces energy costs and improves overall efficiency. Pooling research efforts can also reduce resource consumption.
Giving preference to local and sustainable infrastructure
Deploying models on local infrastructure can reduce data transfer costs between regions. In addition, choosing data centres supplied by energy that is renewable or low CO2 equivalent directly contributes to reducing their carbon footprint.
Consider the lifecycle of models
Responsible management involves the optimization of every stage of implementation: training, deployment, and maintenance. Limiting the stream of data between suppliers or far-flung regions reduces costs and environmental effects. Practices such as resource virtualization and consolidation maximize overall efficiency.
Optimizing with marketing mix modelling
Marketing mix modelling (MMM) is a proven statistical technique that uses historical data to evaluate the impact of marketing activities on a company’s key performance indicators (KPI). Initially developed in the 1950s to measure the effectiveness of TV ads, MMM has recently been the subject of a resurgence of interest. Its new popularity is due to the growing restrictions stemming from privacy protection, which limits the effectiveness of traditional methods of digital tracking, such as multi-touch attribution.
Given a context in which media campaigns contribute to carbon emissions, whether digital, OOH, or TV ads, MMM is emerging as an essential tool for guiding more sustainable decision making. It offers a two-fold opportunity: that of optimizing campaign performance while minimizing environmental impact.
To supplement your reading, take a look at our article on how artificial intelligence and advertising technologies are transforming the marketing industry.
Optimization of marketing budgets
Optimizing the allocation of marketing budgets constitutes an essential step in the integration of environmental objectives into advertising strategies.
MMM is notable for its ability to adopt a multi-goal approach. On the one hand, it aims to maximize marketing results, such as increasing sales or brand awareness. On the other hand, it takes into consideration the carbon emissions associated with campaigns to minimize their environmental impact. By analyzing the performance of various channels (TV, digital, radio, OOH, etc.), MMM lets you identify which media offer the best compromise between return on investment (ROI) and sustainability.
For example, digital campaigns using servers supplied by renewable energy can be given preference, while investments in channels with a more significant carbon footprint, such as television or print, can be reduced. This methodology therefore helps companies reconcile economic efficiency with environmental responsibility.
Selecting a statistical model
The choice of statistical model is a key element in MMM, since it directly influences the precision of analyses and the energy consumption of calculations.
Lightweight models, such as ridge regression or Bayesian models, prove to be particularly well adapted to responding to these factors. These approaches guarantee reliable results while limiting the resources needed for their implementation. In addition, random forest models represent an interesting option for capturing complex interactions between variables while maintaining optimal energy efficiency.
On the other hand, neural networks, while highly adept at modelling non-linear relationships, are often ruled out when undertaking a sustainable approach. The training they require and their execution involve significant energy consumption, which makes them less appropriate in contexts where reducing AI’s carbon footprint is a priority. By giving preference to light, robust models, companies can ensure precise analyses while reducing their environmental impact.
Conclusion
Artificial intelligence offers many opportunities for transforming a variety of sectors, but its adoption needs to be combined with responsible practices. By evaluating the social and environmental impacts of these technologies, businesses can maximize their benefits while reducing their risks. Tools such as MMM allow marketing performance to be combined with sustainability, while generative AI opens up new possibilities in terms of creativity and customization.
By combining innovation, ethics, and responsibility, it is possible to get the most benefit out of artificial intelligence while meeting the social and environmental challenges we face today. Need help with your implementation? Contact our experts in artificial intelligence.
Written by Houssem Bouazizi and Mickael Wajnberg
References
Barroso, L. A., and Clidaras, J. (2022). The datacenter as a computer: An introduction to the design of warehouse-scale machines. Springer Nature.
Ekimetrics. Understand your real marketing carbon footprint to crack sustainable business performance (2/3), https://www.ekimetrics.com/articles/decarbonization-marketing-mix
Gianni, R., Lehtinen, S., and Nieminen, M. (2022). Governance of responsible AI: From ethical guidelines to cooperative policies. Frontiers in Computer Science, 4, 873437.
Husson,Thomas. Toward A Greener Marketing Ecosystem, https://www.forrester.com/blogs/toward-a-greener-marketing-ecosystem/
Strubell, E., Ganesh, A., and McCallum, A. (2020). Energy and policy considerations for modern deep learning research. In Proceedings of the AAAI conference on artificial intelligence (Vol. 34, No. 09, pp. 13693-13696).
Schwartz, R., Dodge, J., Smith, N. A., and Etzioni, O. (2020). Green AI. Communications of the ACM, 63(12), 54-63.
Tremblay, D.-G., Psyché, V. and Yagoubi, A (2023). La mise en œuvre de l’IA dans les organisations est-elle compatible avec une société éthique?, Ad Machina, 7(1), 166-187. https://doi.org/10.1522/radm.no7.1663
