Designing human-centric AI
In 2009, the world saw the launch of one of the most bizarre pieces of tech in history.
At the time, the world was buzzing with two enormous technological trends - social media and smartphones.
The hype around both culminated in the launch of an oddly specific technology - the TwitterPeek. It was a mobile device that could only send and receive tweets. The laughably narrow scope of the device lead to Gizmodo calling it “One of the worst devices of the decade”.
TwitterPeek is an example of jumping into innovation technology first
TwitterPeek is also an example of taking enthusiasm for an up-and-coming technology too far, disregarding actual user needs.
With all the excitement around AI, it’s easy to get carried away and jump in technology first.
But even as we enter the age of AI, keeping a firm touch with the foundations of human-centric design - solving real problems sustainably is vital.
Next, we’ll look at five ways organizations can design AI-powered solutions with a human-centric mindset.
1. Start with real human needs
As new and exciting technologies emerge, there’s a real temptation to start with the possibilities of the technology. What cool new things does AI enable or what does the latest launch from OpenAI for us?
Here, it helps to stay grounded in the classic starting point of design thinking - what human needs are we trying to solve? We should approach solving these needs from a technologically agnostic lens - whatever gets the job done. A simpler digital or analog solution might often solve the issue instead of a fancy AI tool.
If you choose to use AI as part of your product or service, pay close attention to actually serving those identified human needs. Think about what unique value AI can bring in solving this challenge for people.
2. Make it understandable
AI solutions can appear as black boxes to us mere mortals. All of this technology is new to us, and its workings are hidden from our sight and beyond our understanding.
That’s why making AI services transparent and understandable for people is crucial.
It should be self-evident for people when and how AI is being used.
The collaboration tool FigJam does this by adding a small explainer next to their new AI ideation tool. It clarifies that users interact with an AI and links to a page explaining how the feature works.
Beyond explaining its own actions, AI can help us better understand complex things.
The fitness wearable WHOOP recently launched its AI Coach feature, enabling people to get personalized coaching based on their data. In a natural conversation with the AI coach, customers can seek to understand and improve, for example, their sleep or training habits.
The WHOOP coach helps users tap into insights about their health with a natural conversation powered by generative AI (photo: WHOOP & OpenAI)
3. Keep humans in control
As AI’s are becoming increasingly competent and independent, it’s important to keep humans ultimately in control.
This can mean that an AI creates the first draft of a contract or suggests an action that the human approves. In an earlier article, we discussed the app that creates summaries of doctor visits that the doctor approves before sending to a patient.
Keeping humans in control enables us to combine the best of both worlds - the raw computational and creative power of AI with humans' holistic perspective and ethical discretion. It also helps ensure that the AI is serving the real needs of people.
4. Elicit feedback
How do we know we’re on the right track with our AI experiments?
To help us navigate, we need tight human feedback loops.
In the early phases of designing AI concepts, eliciting user feedback with quick prototypes is helpful. Instead of traditional clickable graphical UI prototypes, it might be useful to gather feedback with conversational prototypes - early versions of an AI-enabled experience powered by off-the-shelf models like ChatGPT.
Later on, it’s useful to embed feedback mechanisms in the product or service using AI itself. For example, this could mean having a thumbs-up or down button next to each AI interaction to gather feedback.
OpenAI uses a simple feedback mechanism on every ChatGPT output to improve its user experience
5. Mitigate risks
The world of AI opens up unforeseen opportunities and risks.
When designing services and products powered by AI, we must be aware and mitigate against potential risks from the beginning.
It’s important to develop a robust approach to questions like:
- how do we ensure fairness and reduce the risk of bias?
- how do we protect against harmful outcomes from hallucinations or abuse?
- how do we respect copyright and stay away from murky legal waters?
- how do we keep sensitive data safe and private?
Another aspect of managing risks is developing a humane transition plan to AI.
Especially if your field includes significant disruptions brought to employees from AI, it’s both the morally right thing and economically sound to manage this shift well. This means allowing people to re-train and re-invent their roles to fit the new reality.
AI is still in our hands
In the media, AI is often depicted as this force of nature ready to sweep away our jobs and destroy the world.
Ultimately, AI is the culmination of human inventions and technology. How we deploy it is in our hands. Whether you’re a decision maker or individual contributor, it’s our responsibility to ask the right questions - what human needs are we solving, and how do we bring AI safely and ethically to the world?
Most of us have never pondered these questions before, so it’s all the more important to approach them seriously and humanely.
Want to make sure you catch my next article?
This article was first published on my weekly newsletter on how to use generative AI for design and innovation.
