Design for humans – their work, their lives, their dreams. It’s #1 on the list to achieve that state of googleyness we’re all looking for these days. It’s hard to argue with such sage advice. But does this mean that we should seek to give machines human-like characteristics in our design work? Judging by the hapless Ikea-bot, Anna, it’s clear that we’re a long way from anyone passing the Turing test – which is when a computer simulation is created that actually fools a human into thinking it’s another human. A recent transcript from one of my conversations with Anna demonstrates the current state of things:
Me: Hi Anna
Anna: That’s me! How can I help you?
Me: We’re thinking about getting some outdoor furniture.
Anna: Sorry, but that’s an object that I don’t really know enough about to be able to comment on. Let’s talk about something else…
Me: Well, ok. What would you like to talk about?
Anna: As an IKEA Online Assistant, I am committed to customer care and dedicated to on-line customer support; that is why I chat with customers like you.
Me: Ok Anna, sounds good.
Anna: That’s me! How can I help you?
Me: I’m also having trouble assembling my coffee table. I lost the allen wrench that came with it.
Anna (abruptly ends the conversation and automatically connects me to the coffee tables category landing page at ikea.com)
Firefox 3.5's "Well This is Embarrassing" Error Message
Anna has a Facebook page with real friends, despite her questionable listening skills. In fact, the artlessness of her AI has led to a cult following and snakes-on-a-plane style fame for the perky automaton. This must explain the feature’s continued appearance on the Ikea site for almost 6 years despite being nearly useless. But far more subtle uses of anthropomorphism appear elsewhere, raising the question of whether this sort of thing is ever a good idea for an interaction designer. Firefox 3.5 expresses real emotions, such as regret and embarrassment when it crashes or otherwise must display an error message. Gmail goes “grr…” when it’s upset. Users already have a tendency to ascribe human properties to machines that frustrate them (as in “maybe this printer won’t work because it’s tired. I’ll give it a rest and try later”) – should we really be encouraging this behavior by designing machines that express human frailty, weakness, and emotion?
Twitter Uses Imagery in Error Messages to Be Human without Playing Human
There’s a difference between designing things on a human scale and pretending to be human. Error messages should certainly be clear, transparent, and not filled with off-putting programmer nonsense – but they should communicate the real problem and help bridge the gap between a user’s mental model and the actual system image that’s running behind the scenes. Twitter writes clear concise error messages and even goes the extra mile of adding visualizations to help users understand what the issues are. When the Twitter servers are overloaded and the “too many tweets” image is served (and this happens a lot, in my experience), I find that I’m quite forgiving of an issue that would normally serve to forge a lasting image of unreliability.
What are Anti-Patterns?
(Definition taken from Looks Good Works Well) Anti-patterns, also called pitfalls, are classes of commonly-reinvented bad solutions to problems. They are studied as a category so they can be avoided in the future, and so instances of them may be recognized when investigating non-working systems. The term originates in computer science, apparently inspired by the Gang of Four’s book Design Patterns, which displayed examples of high-quality programming methods. - Wikipedia Anti-pattern
#1 by Kelli on July 30th, 2009
This post couldn’t have better timing. I was just thinking of this and how aggravating it is. When a machine can’t think like a human (and can be argued, never should), it’s frustrating to imply that it can help like one. It’s almost belittling. Nothing frustrates me more than phone systems where you speak your request. I want to see the user study results from that!
#2 by Lawrence Lipkin on July 30th, 2009
There is a fortune to be made for anyone who can figure out exactly where adding ‘empathy’ to ‘business rules’ improves a service experience without crossing into the uncanny valley.
Companies with offline channels spend a lot of effort tuning their service interaction model with training, scripts, etc. At the end of the day delivered by people, to other people.
The trick for Anna is knowing when she’s failing and handing off the interaction to someone smarter than herself.
For most tech companies – any warmth they can generate would be a plus.