Summary: AI desperately needs UX. Why? Ethical AI legislation emphasizes the need for your AI to adapt to keep AI UX relevant in a user’s task flow. Generative AI will require a human-centred approach that is ethical and inclusive by default.
Why does AI desperately need UX? 85% Fail?
You may have seen this statistic:
85% of artificial intelligence (AI) and machine learning (ML) projects fail to produce a return for the business.
The reasons cited for the high failure rate included poor scope definition, bad training data, organizational inertia, lack of process change, mission creep, and insufficient experimentation (Forbes, Gartner, Inc. estimate).
However, it is essential to note that those estimates were made before generative AI–which was released widely in early 2023 (ChatGPT, etc). Nonetheless, by learning from the recent past, we can better prepare for now and the near future.
A 2017 study in the UX community revealed an additional issue with AI and UX: UX professionals reported not being involved in Machine Learning (ML) projects or being unprepared for the challenge of prototyping “invisible” interfaces. Worse, UX designers were invited at the end of a project-– when influencing positive UX outcomes is too late.
[UX designers] described difficulties in prototyping ML, expressing ML ideas, and noted the need for designers to collaborate with skilled technologists. One noted, “Machine learning is hard to prototype. Machine learning requires highly skilled collaborators, when a lot of companies are only able to hire ‘warm bodies’.” Additionally, another respondent stated, “…making interactive prototypes that incorporates machine learning is hard (haven’t found a way to do that yet in an easy fashion),” and another indicated that UX designers make statements such as “ …inputs come in …some magic happens …and all your business needs are met!” Clearly, in order to effectively leverage ML as a design material, designers currently feel that collaborations are essential.
Clearly, there is a critical need for early-on collaboration among engineering and UX teams around AI experience. But it’s not just about making a relevant and appealing user experience. It’s about safety.
Make your AI safe, or face legal risks.
The latest AI regulation, signed in late 2023 by China, the US, and the EU is called The Bletchley Agreement. It spells out future directions for AI:
“[AI will be] Human-centric, trustworthy, and responsible.” The focus of the new international agreement is on AI safety:
“There is potential for serious, even catastrophic, harm, either deliberate or unintentional, stemming from the most significant capabilities of these AI models.”
So it makes sense that all AI projects should have a focus on UX and what the AI community calls FATE (Fairness, Accountability, Trust, and Ethics). Microsoft has a group dedicated to FATE.
A recent study from Microsoft FATE researchers (2023) found almost identical issues to the earlier study mentioned above: UX folks need to be involved in AI work, and there are challenges:
Our study highlights the pivotal role that UX designers can play in Responsible AI and calls for supporting their understanding of AI limitations through model transparency and interrogation.
Helping users get more from AI interaction design
Zendesk’s CX Trends 2023 report found that consumers have a clear idea about what they want from AI and customer service interactions:
Consumers want AI that creates more personalized and effective customer support. For example, AI that curates an outfit for a consumer based on prior shopping habits. About 2 out of 3 of consumers would be willing to share more data to have a richer experience with AI.
On the other hand, Most consumers would be less inclined to use an AI/chatbot if it failed to provide more personalized experiences.
However, another study among young people found increased concerns about privacy UX:
62% of 18-24 year-olds “expressed their concern about how organizations could be using their personal data for AI.” (Cisco study 2023).
AI should demonstrate empathy, according to consumers. About 7 out of 10 think AI should correctly understand and respond to their emotions and feelings.
As AI implementations grow exponentially, UX issues must be addressed.
In e-commerce, AI adoption is increasing exponentially, according to the Salesforce 2023 Connected Shopper study:
*92% of retailers say they are investing in AI more than ever to improve shopping experiences
*17% of shoppers say they’ve used generative AI to get inspiration for product purchases.
UX in AI solution development means resonating with the users’ needs, desires, and behaviors. All Interaction Design comes from this due diligence activity. UX professionals can craft AI solutions that integrate seamlessly into users’ lives by understanding the intricacies of current behavior: users and pre-trained language models. These efforts should enhance the AI user experience and default to safety and inclusion.
Meta is about to unleash generative AI experiences across its platforms. To avoid downstream issues, such as public relations from bad UX, they conducted eight months of adversarial testing or Red-teaming:
For Meta’s latest rollout of AI chatbots, they spent 6,000 hours red-teaming the model to find problematic use cases… (Alex Heath, The Verge, Sept 27, 2023).
Why UX for AI now?
First, you want to avoid AI features that lack practical value. As the market inundates users with an array of AI-driven products and services, the allure of adding AI features for the sake of novelty can be tempting.
Next, it’s vital to de-bias AI data set training or provide balancing mechanisms to weed out poor, irrelevant, or biased results or outputs. One of the remarkable UX features in ChatGPT is the feedback > aut0generate result interactions. First, you give feedback. Next, it rewards you with an alternate result. That interaction is unique and unlike any user experience created to date.
Finally, basic usability is missing from how users access the current generation of Generative AI tools. Problems include sign-up (think MidJourney), log-in, or re-visiting the tool (e.g. Bing DALLE-3), lack of search (ChatGPT), and more. Don’t overlook these novice usability mistakes, or they will hurt your user adoption.
Biggest challenge ahead
One of the biggest challenges will be establishing relevance in user’s lives. Worse, integrating AI without a robust UX strategy can lead to a disconnect between user expectations and the actual capabilities of the AI system. Users may have inflated expectations of your AI, anticipating solutions that comprehend their needs intuitively and offer personalized experiences effortlessly. When these expectations are unmet, users can quickly lose trust in AI-driven solutions, leading to skepticism and reluctance to adopt future advancements.
Beyond the Tech community, re-visiting an AI tool and making it indispensable to users’ work challenges or life events will differentiate long-term adoption. Many general population users tried ChatGPT for example, then stopped using it after a few visits.
Key Take-Aways:
AI needs UX fast. Businesses often overlook the significance of ethical considerations and the impact of AI on society at large. Integrating AI without ethical foresight can perpetuate biases, reinforce societal inequalities, and infringe upon user privacy. UX professionals with a solid ethical framework are pivotal in steering the AI gold rush toward a path of responsible innovation. By advocating for inclusive design principles, ethical data usage, and transparent AI decision-making processes, UX professionals can ensure AI user trust and contribute to safer and higher adoption of AI
Final Thoughts
The collaboration between AI and UX is not a one-time affair but an ongoing dialogue that necessitates continuous iteration and refinement. Through user research, usability testing, and iterative design processes, UX professionals can gather invaluable insights that inform the enhancement and optimization of AI-driven solutions, fostering a user-centric approach that remains adaptable and responsive to evolving user needs and preferences.
Learn more: AI for UX Masterclass.
