The human characteristics of the best software don't come from AI
AI isn’t inherently human-centred, effective interfaces still require empathy, intuition, trust, personality, anticipation and forgiveness.
40 years ago, the discipline of Human-Computer Interaction emerged, championing the simple idea that successful software needed to consider the human as deeply as the computer to be successful, and that technology by itself could never offer a magic bullet.
It is ironic therefore that today we appear to be unwiring 40 years of hard-learned lessons by increasingly championing AI as the silver bullet solution to humanising technology.
And that’s before we get to the many accuracy, ethical and environmental questions which AI raises.
An examination of the characteristics of human-centred software highlights why it is so wrong to assume that AI comes with humanity built it. It doesn’t. It barely gets close.
All user interfaces, whether graphical, voice or text-based, have a staggering amount of work to do to achieve their goal of letting humans and computers communicate effectively in two directions. Their job is to get the emotional error-prone ethical inferential intelligent irrational random slow unpredictable human to interact with the amoral, apathetic, dumb, error-free, fast, literal, logical, predictable, sequential computer.
Humanised interfaces therefore seek to reduce friction in the interface, by infusing software with human-like characteristics, such as the following:
It shows empathy
Software which displays empathy understands its users deeply, allowing it to answer their questions, allay their concerns and reinforce their positive motivations as part of the delivery of functionality. When a software-as-a-service website tells me in advance that it won’t need a credit card to allow me to start a free trial, it is recognising that this is one of my questions, and a barrier for me to use it, and is allaying the concern before it grows. When my email client gives me an “Email sent” status update it is assuring me that the email has gone, meaning I don’t need to check my Sent items to make sure.
It is personal
When I go to a favourite restaurant, I appreciate it when the maître d remembers my name, enquires about my family and comments on how my favourite rugby team is performing. So too, personalised software is invested in understanding its users, their preferences and previous behaviours. When a website loads information on its home page specific to my preferences, or sends me an email in advance of a birthday or anniversary, it is making its functionality personal.
It anticipates
It is said that really good friends can finish each other’s sentences. Software with good anticipation uses my past and present behaviours and interactions to predict my future needs. When my online form pre-fills on loading, or typing the first few letters of an email address results in the remainder being optionally auto-completed, the system is anticipating my needs, shortening my task time and making things easier for me.
It is intuitive
Good software works just as I expect it to. Over a quarter of the century after the invention of the world wide web, successful and effective design patterns have emerged which dominate common tasks such as software onboarding, e-commerce purchasing and even navigation and search. Intuitive interfaces simply work the way the user expects. When the check-out flow on the new e-commerce website asks me for information in the order I expect, and has Google Pay integrated, it feels easier than other ways of buying.
It displays awareness
Humanised software is constantly using contextual cues to provide a better experience for the user. When the web page I am viewing renders differently depending on the device I am viewing it on, it is exhibiting awareness. When Apple Wallet displays my boarding pass as I approach the airport, it is displaying responsiveness. Like humans, good software avoids a rigid approach to decision-making, and uses a more fluid, flexible, contextually sensitive one to help deal with the unpredictability and occasional chaos of the real world.
It exhibits personality
The best software demonstrates a confidence around its tone of voice, both written and visual, and speaks with its users in a language which they understand and appreciate. There is no correct approach to this – the tone of voice might be serious, or workmanlike, or playful – but as each human is unique so too effective software takes seriously the tone in which it communicates. When Slack launched with an Emoji library, it was clear they wanted to encourage creative and rich communications. When it subsequently expanded that library to allow users to upload their own Emojis they were recognising the importance of personality in communications. When gov.uk determined that they would write their website content based around a reading age of 9, they were recognising the importance of understanding in their interface’s success and helping the greatest number of their users access UK public services.
It engenders trust
Trust-based collaboration is at the heart of human survival and advancement. It’s therefore no surprise that deserving, earning and communicating trust is a key part of effective software. Trust signals such as third-party reviews, case studies, customer success stories, and customer testimonials are a key part of website persuasion, but trust goes beyond this. In software, users need to trust the system does what it says it will and need to feel confident that they won’t break anything if they interact. Clear labelling and navigation, progress indication, and constant communication of system status all contribute to helping the user feel confident and in control of a system they can trust.
It forgives
Good software knows users will make future mistakes and have made previous mistakes. It therefore tries to prevent potential mistakes and remedy previous ones. Sometimes software asks “Are you sure” when you are about to do something important, irreversible, or that costs you money. Wizards guide users through difficult tasks when they don’t want to steer themselves. And not uncommonly the Undo button is our best friend. Sometimes though, forgiveness can be softer, like when I pause my podcast, and when I press play it starts 10 seconds behind where I paused, allowing me to get gently up to speed with where I was.
It’s not that AI interfaces can’t exhibit these qualities, it’s just that they don’t come out of the box. The LLM and Gen AI engines which power the world’s leading AI interfaces are enormous repositories of content brimming with potential, yet just as infants need to grow to adulthood, so AI has some distance to travel before it meets the standards of best-in-class HCI.
