• Digital
• Thought Leadership

Sustainable Design Processes and the AI feedback loop

12 months ago by Alice Thompson

Following on from our Design & AI Autumn Conversation events, Alice Thompson (Industrial and UX/UI designer) shares some thoughts with us on the AI feedback loop and its affect on sustainable design processes.

To see the recap from the Design & AI Autumn Conversations events click here.

Interested to learn more about integrating AI into your design practice? Come along to the workshop, ‘Taking advantage of AI to elevate your design practice’ being held in Auckland and Wellington in June 2023.

---

The sheer pace at which Artificial Intelligence (AI) technologies have appeared in creative industries over the past couple of years and continue to advance makes any commentary on the future of AI and design feel speculative and subjective. But, with a tool like AI rooted in efficiency and optimisation becoming increasingly a part of the design process, how do we ensure the values of the tool don’t compromise those of the process itself? 

Artificial Intelligence is already being used in various stages of the design process. Generative AI tools such as Midjourney, DALL-E and Stable Diffusion are able to convert text inputs into high-quality AI-generated images, shifting creative and conceptual practices. Others such as ChatGPT-4 are able to help with time-intensive manual tasks such as collating and synthesising quantitative research. Many of these softwares are barely over a year old but are already being hailed for their ability to streamline design processes and have sparked discussion as to the implications on the future of design.

AI is also showing great promise when it comes to sustainability, with the technology holding opportunities to create designs that are lower in carbon emissions, embodied energy and wastage. Similarly, with its ability to hybridise multiple concepts, AI can help us further explore the intersection of nature and design through ideas such as biomimicry and biophilic design.

But sustainability in design encompasses more than just energy usage and ecological implications. It also considers the further complexities concerning social & political equity and inclusivity, all of which stem from within design processes. 

A commonly talked about challenge with AI technologies is the issue of biases. As AI technologies draw from collections of data, whether that’s collated training data or the collective internet itself, generative or analytical outputs from these tools are only going to be as inclusive and diverse as the pool of data they stem from. Being interpolative in nature, AI usually takes the average of what it draws from to generate an output. So when the data is informed from sources that haven’t held the same space for marginalised perspectives or experiences, for example, there is a risk of homogenised outcomes that perpetuate incorrect or harmful information. 

Furthermore, valid concerns surrounding the diversity of those involved in the design and testing of the AI systems themselves are another issue when it comes to sustainable design with AI. Just as the scope of data can provide limitations, so can the algorithms that process them. Joel Maxwell discussed the differences between te reo Māori and English when describing his experience in speaking with new AI, “In English there can be 20 words for one thing while reo Māori often has one word for 20 things…it means in te reo, context is paramount to meaning” (Maxwell, 2023). 

Nuances in languages and concepts are already difficult to accurately capture in a quantitative manner. For those of indigenous groups, this is even more so when the AI systems these nuances are expected to fit into are tested and designed generally without involving those whose knowledge and experience they belong to. So again the outputs that may influence the design process are only going to be as inclusive, as diverse, and as sustainable as the latent space allows.

Current AI approaches, at their core, centre around optimising a particular measure as a target. What is metrically deemed successful or accurate can also have implications on sustainable design outcomes if the success of the tool itself (as defined by limited metrics) takes priority over the quality of the design process.  

As the landscape of design shifts, merging the physical more and more with the digital, it will naturally create a feedback loop that sees the designs created determine what gets created next. If these, however, have been influenced by algorithms whose metrics for success are defined in an unsustainable way, whatever design outcomes follow will prove potentially problematic.

Although it may not sound like it, I am excited about the future of AI and sustainable design practices. As this technology advances it’s a reminder that whatever direction it will take comes down to us. While AI is not in its infancy, it is at a place in relation to creative industries where the habits and practices implemented when using it are still malleable. 

Dropping AI tools within existing frameworks to replace old processes for the sake of efficiency will only see optimisation to a limited extent. Instead, let’s design processes that are considered and take into account AI’s shortcomings.

Just as scientists document and reflect on the limitations of their trials, such as underlying assumptions and influencing factors, a similar awareness and level of scrutiny should be applied when utilising AI tools within design. A symbiotic relationship between AI tools and an industry responsible for creating products and services that directly impact lives and communities must actively work to maintain emotional design intelligence and integrity.

Innovation lies in the outliers. So does great design that is lasting and founded on experiences that are true to those they hope to impact. Neither can occur through processes that are solely driven by models where success is centred around metric optimisation. If we’re wanting designs that are sustainable and inclusive then we need to be intentional in the processes and tools we use to create them. Sustainable design processes that include AI tools are possible, we just have to take the time to design them.

References