Participatory Design approach to create AI for sustainable digital futures: Individual Essay

Introduction  

We live in a contemporary landscape of ever-emerging technologies and innovations that are introduced to the masses at a very fast pace. These technologies fall within the realm of Industry 4.0. The term ‘Industry 4.0’ was introduced in the year 2011, in Germany (Paksoy et al., 2020) to describe the era of a Fourth Industrial Revolution. Built on the fundamentals of Industry 3.0’s Internet, Industry 4.0 takes it a step further merging the Physical, Digital and Biological to create technologies that transcend the ways in which industries and individuals function (Paksoy et al., 2020). Industry 4.0 technologies are categorised into four main typologies by (Anon, n.d.): 1. Connectivity, data, and computational power 2. Analytics and intelligence 3. Human–machine interaction 4. Advanced engineering. These widespread categories not only affect the way we work but also heavily dictate how individuals live and how societal shifts take place.

At the core of this paradigm shift is Artificial Intelligence (AI), representing the analytical and intelligence pillar of the Industry 4.0. Artificial Intelligence as a term first coined in the year 1942 by Isaac Asimov published his short story Runaround. It became an established theoretical concept in 1956 (Haenlein and Kaplan, 2019). Runaround as a piece  of writing can be seen as the birth of AI, similar to many other digital innovations that have their roots in science fiction. AI today, has become widespread, influential, and important within our daily lives permeating through every industry (Wang et al., 2024). AI was created as a tool to extend human intelligence (Kaplan and Haenlein, 2019) and aid humans to conduct their tasks in a faster and more effective manner. The intent was to increase efficiency and economic expansion of the value chain of a product or service. Although, we are witnessing negative affordances of this innovation. These negative impacts include carbon emissions, potential substitution of human jobs, biased AIs, and diminishing creative integrity. It is imperative to guard against creating long lasting negative impacts of AI. To do so is easier said than done, as each technology brings with it challenges that arise parallel to its development. As mentioned earlier, the fast-paced nature of innovation can leave the masses to play catch up with the technology. Furthermore, it could leave gaps within the creation of the technology to analyse its impacts and the affordances it creates.

To combat this, inclusions of stakeholders that make up the demographic that is potentially going to use or be affected by the AI would be the first step towards breaking the barrier between created for and created with. A Participatory Design approach gives us a pathway to facilitate this inclusive creation method. It will enable us to bridge certain gaps that AI has been creating, for example, Racial Biases. Including the users within the creation process would allow the developers of the AI to view the product from multiple perspectives instead of one limited to the creator. A Participatory Design approach would simulate conversations between various stakeholders, creating AI that is informed, equitable and inclusive. Subsequently, this would lead to a sustainable development of AI for Humans. Sustainability can be justified as the resilient interlink between Humans and the environment (Harré et al., 2022). Sustainability can often be viewed from a skewed lens of Humans being responsible for sustainability. In its truest from it is a valid statement, our rapid man-made creations are threatening the welfare of our ecosystem. However, we often forget that humans are an innate part of the ecosystem itself. Participatory design can stimulate a people-centric system to tackle sustainability issues that are rising due to the widespread adaptation of AI. In this essay I will explore the measures we can take to create a sustainable AI through Participatory Design, for it to act as a viable agent for humans today and our future generations.

Current landscape of Artificial Intelligence

Artificial Intelligence in its essence can be described as “the science and engineering of making intelligent machines” (Manning, 2022). This definition is valid; however, it is a rigid technical description of Artificial Intelligence. IBM describes AI as “a technology that enables computers and machines to simulate human intelligence and problem-solving capabilities” (Anon, n.d.). This humanises artificial intelligence, making it a collaborator to humans instead of simply being a tool one uses (Korteling et al., 2021). Artificial intelligence is an algorithm trained using a data set through Natural Language Processing to perceive, learn, make decisions and problem solve. AI can be further categorised into two types: Narrow AI and Deep AI. Narrow AI is created to perform a specific set of tasks, trained on a specific data set, for example, The Louvre Chatbot to specifically interact with respect to the exhibits within the museum. Deep AI is a holistic approach, enabling the AI to mimic more comprehensive Human based tasks (Kumari, n.d.). One such mainstream example is ChatGPT, trained on extensive amount of information such as the internet, books, journals, articles, websites and more. 

Delving further into the applications of AI: it is being used in significant industries such as Healthcare, Law, Education, Defence, National Security, Finance, Engineering, Retail, Administration, Autonomous Vehicles, Biometric technologies, supply chain management, social media and potentially more. It is evident that the applications of AI are permeating into every industry and some of the societally significant ones at large. The quality of human to take empathetic decisions that are governed more than by just facts are lacking within the AI technology. This particular lack of empathetic ability, specifically emotional empathy is what makes AI “Human like” and not Human (Montemayor et al., 2022). This is a significant threat when we consider AI being used within Law, Healthcare and Defence. Let us take an example within the healthcare industry, Google’s AI that is being trained on Mammograms to detect breast cancer. The AI was being trained on a significant data set, it proved to be reduce false positives by 1.7% and false negatives by 2.7% (Khalid, 2020). The numbers project a positive stance on implementing such an AI, but what was left unchecked was taking into account the diverse races that should have been integrated within this AI. Similarly in the book ‘Data Feminism’, this societal issue is explored through the lens of Black Women being more susceptible to death during childbirth because the healthcare systems weren’t designed for them or with people like them in mind (D’Ignazio and Klein, 2020). 

This is one of many examples that informs us about the biases and disparities within our society. Artificial intelligence is trained on data that pre-exists. But what about when this data is tainted with our hegemonic ways? Ethics for AI do exist, but they are privy to the same biases and cannot actively keep up with the dynamic nature of updating technologies. The issues that we face with AI today at an interpersonal level can only be addressed when we address the way our society functions. When we question who is creating these technologies, what data set is informing them and do the people get a say in how they want the AI to aid them. Introducing Participatory Design approach to include relevant stakeholders to create actively informed and ethical AI. This will be an efficient way to negate these issues that AIs today generate.

“If we were concerned about developing ethics for humans using machines, now we urgently need to discuss ethics for machines.”

-        (Bertoncini and Serafim, 2023)

Sustainability and Artificial Intelligence

With each Industrial Revolution, the world has adapted to newer technologies that require rising power consumption. These technological adaptations traverse through variations with each industrial era. From steam engines to the printing press, to mass production, to the internet and now to AI, IoT and Machine Learning. The AI industry that was valued at $100 Billion in 2021 is estimated to expand to a staggering $2 trillion by 2030 (Wang et al., 2024). The intent behind AI is streamlined economic development structures, effective Industrial work-flows and advancement, rewired energy frameworks, thus leading to a sustainable future (Wang et al., 2024). All these goals that were set for the implementation of AI are facing a paradox. The paradox can be explained in terms of the energy consumption required to train and run an AI is superseding the sustainable goals it was meant to achieve. Furthermore, AI carbon emissions can be mapped through the “Rebound Effect” (Wang et al., 2024). The rebound effect describes the development of new energy saving technologies. These technologies can significantly reduce the cost of the energy being used, which could be considered as beneficial in terms of lower energy bills (Brännlund et al., 2007). However, the lowering of the cost of a good such as energy will lead to an increase in the consumption, opposing any benefits that could have brought to the forefront (Brännlund et al., 2007). I believe this phenomenon does not grasp the degree of energy consumption and the need to develop newer energy standards to run large scale AIs. The hardware required to train and run an AI is developing exponentially. The industry is playing quickly implemented catchups to keep up with this trend. Are these rapid developments holistically adhering to the sustainability guidelines we have in place?

AI for Sustainability

            Climate change is a threat that will lead us to a dystopian future if not tackled with the right means and at the utmost urgency. AI is being used as one of the tools to address this crisis in hand. AI can analyse a vast multi-dimensional data set and generate trends and forecast when trained with the right set of optimisation tools (Cowls et al., 2021). Currently, AI is being used to predict global temperature changes, map cloud systems, rainfall assessment and better understand other weather phenomenon (Cowls et al., 2021). Artificial Intelligence is also being implemented to forecast major-disrupting climatic events, to set forward informed Disaster Management programs in place. Application of AI within this realm could potentially save us or efficiently equip us to brave these significant calamities.

AI against Sustainability

            In the topic ‘Current Landscape of AI’ we explored the Biased and societally unsustainable implications of AI. In this section we focus on unsustainable impacts of AI on the environment. AI for climate predictions and trend forecasting is trained on an extensive data set dating to the origin of the recorded data. This process of training the AI is quite taxing on the environment. Let us take the example of ChatGPT 3, it “consumed 1,287 megawatt hours of electricity and generated 552 tons of carbon dioxide equivalent, the equivalent of 123 gasoline-powered passenger vehicles driven for one year” (US, n.d.). This does not include the resources required to run the AI that are recurring and significant. ChatGPT releases newer versions, training it on newer data sets. This is a never-ending source of carbon emissions, considering our AI technologies are only going to advance with the times. Now, studying the hardware built and used to access – run the AI, they will eventually end up in a landfill to add to the existing piling of E-waste (Haakman et al., 2021). Albeit the future development of AI is meant to make it more sustainable and holistic. There will have to be a significant gap between the emissions that AI produces versus the sustainable efficiency it creates to build a Net Zero world. As for achieving sustainable climatic outputs, Policy making will play an influential role to regulate the negative effects of AI whilst feeding its innovative edge.

            The recognition of these threats that AI poses has led to the concept of ‘Green AI’. Green AI is the research to analyse and moderate the negative effects of AI and unlock its true potential to design sustainable impacts (Verdecchia et al., 2023). (Verdecchia et al., 2023) analyses the current understanding of Green AI within research and industry professionals. It summarises that the field of Green AI is growing but the users do not have a holistic understanding of the field. Furthermore, it highlights the limited involvement of Industry experts that are privy to this space of knowledge (Verdecchia et al., 2023). This shortcoming leads to uniformed AIs within specialised industries, thus, snowballing into biases or unreliable outputs. I concur with this point of argument, AI as a concept is widely known but the industries are not equipped enough to grasp the intricacies of creating one. This furthers the application of Participatory Design approach, to include who we are designing the AI for and create with them. This includes the Client and the end user.

Participatory Design to create Artificial Intelligence

            Participatory Design (PD) was first spoken about in 1971 at a Design Research Society (Luck, 2018). Participatory Design (PD) approach challenges the involvement of technologies in workplaces, homes, societies, and government institutions (Muller and Kuhn, 1993). It is a democratic approach that involves the end users during the process of Design to create an artefact (in this case AI) that is most viable for them (Medici, 2020). PD fosters a culture of “we” and mutual learning to build what actually counts (Gharat, 2023). This adaptive way of working leads to innovative and inclusive solutions that engage a diverse range of stakeholders, fostering sustainable innovation (Gharat, 2023).

PD has the potential to address ethical concerns rising due to AI in the realm of biases, transparency, and accountability. As defined above, PD includes various stakeholders, namely, users, policy makers, governance bodies, industry experts, affected demographics and more. The involvement of this diverse personas can help designers identify potential biases and shape well-informed data sets to develop an inclusive and ethical AI. On the other hand, this approach would create transparency within the users about the technology and its functions. Transparency and knowledge about AI’s creation can ensure the reduction of job displacements. Users would adopt AI as a tool or a hybrid co-worker instead of feeling threatened by the technology (Kaplan and Haenlein, 2019). This forms the fundamentals of co-realisation; it is described as mutual learning in its essence (Medici, 2020). It creates a sense trust and acceptance towards the innovation, in turn making it more adaptable and ensures seamless user experiences.

            At a global sustainability scale, Participatory Design could set the interaction between Policy Makers and Designers. This direction would regulate the existing sustainability norms to create efficient AI, which will have a higher chance of meeting a Net Zero outcome. Whilst existing standards can be reinforced, there will be opportunities for this collaboration to foster new laws and regulations that align with the changing innovation. Doing so will enable dynamic policy making that keeps up with the pace of the ever-expanding technology.

Every innovation is created to nurture set affordances. Nonetheless, every such innovation will also have hidden affordances that may or may not be perceived in a positive way (Chen et al., 2009). Participatory Design can help navigate this uncertain territory with a holistic design method that includes all the stakeholders. It would inform the designers about a larger set of user perspectives, mitigating any assumptions or pre-conceived notions a designer would have had to take. These methods could include: 1. Exploratory sessions – preliminary workshops and brainstorming. 2. Shared Research – co-realisation of the technology through the designers and the needs from the users. 3. Discovery Process – clarifying the goals and values of the desired outcome of the project. 4. Prototyping – Iterating the technology to reach the said outcomes (Spinuzzi, 2005). These methods can be conducting through various mediums. It is hard to define them, as they are unique to each project. Overall PD can create valuable frameworks for the development of AI technologies. It will play a key role in ensuring sustainable and value driven deployment of AI into our society.

Case Study

A small Portuguese town named Amiasis was part of the LOCUS project to explore its culture, heritage, traditions, and architecture (Gonçalves, et al., 2022). This exploration was done with the intent of creating a Metaverse based Amiasis to preserve the heritage of the town. The Metaverse captures the intricacies of the town’s architecture and goes into the depths of recreating the characters inspired by the locals (Gonçalves, et al., 2022). The aim of this study was to preserve the heritage digitally and create awareness about the town.

This was done through a Co-design workshop, where the researchers interacted with the residents, engaged in various activities, and conducted interviews. This enabled them to study the demographics, constraint, and biases. All this was done through the Town Council with due information and permissions. It is now a live metaverse with multi-user engagement for educational purposes (Gonçalves, et al., 2022).

Link to the Metaverse: https://maps.secondlife.com/secondlife/AMIAIS%201/91/200/37

Even though this case-study investigates a different type of innovation and a co-design method, the parallels with AI through Participatory design are undeniable. It included the user and other relevant stakeholders to develop an informed innovative technology. The sole purpose behind this project was the sustainability of the heritage of the town. It was a successful project, arguably free of biases and constraints. The only potential drawback was the time and participative commitment required by the researchers.

Image 1 A recriação de Amiais no metaverso Second Life. Locus Social. YouTube.

Limitations

Since Participatory Design includes various stakeholders for the development and informing of AI, it could be seen as an “evolution, not revolution” (Sumner and Stolze, 1997). This could slow down the generation of AI technologies. It would require additional amount of resources such as time, money, research and institutional corporation (Spinuzzi, 2005). Furthermore, the ambiguity of which methods to follow can lead to counterproductive and repetitive PD systems. Participatory design will have to be intricately thought of and curated for project specific needs. Will the investment in the time and resources to conduct PD significantly trade off the sustainability and ethical challenges of AI? It potentially will, by creating streamlined system with minimal interchanges for project-based requirements.

Conclusion

This essay explores Industry 4.0 and how AI is a pivotal technology budding in this era. It delves into the nuances of AI as a technology, what it means and its widespread applications. These applications have direct implications on the society and the environment. As AI continues to reshape our Utopian landscape, industries, and society, it is essential to navigate sustainability and the ethical considerations that come with. AI as a technology will keep developing and take a more sustainable form overtime. Although, as facilitators and users of AI, we need to mold it into a holistic innovation. This can be done through the medium of Participatory design, offering us a favourable avenue for inclusive, ethical, transparent, and sustainable AIs. Regardless of the challenges we might face to deploy these approaches, we can use them to create AI that adopts well-being, shared intellect, value added propositions and environmental integrity. To conclude, Artificial Intelligence, Participatory Design and Sustainability are strong well-founded systems in isolation. The appropriate interdisciplinary approach would equip our world with powerful tools that will adapt to every aspect of living within a dynamic ecosystem.