If interaction design really is the business of behaviour change I believe this must apply two ways. While it’s true that design can influence users and engender cultural change, this is always a product of our more tangible work: changing the behaviour of technology. As a user-centred designer of technology my goal is simple: to make its behaviour humane. But how should I approach this?

Humanity implies emotion and, beneath that, personality. These areas lie beyond the frontiers of classical HCI and usability. Fortunately, as often happens, we view the distant summit and see others have already planted the flag. Toymakers, for instance, have explored the art of bestowing personality on products for years. The results are fairly crude, but I defy anyone to watch the torture of a Pleo and successfully suppress a twinge of guilt. Even in its moments of crisis, Pleo has a distinct personality; that is to say, it conveys emotional information

Channels for personality

Perhaps the most obvious conduit for emotional content is appearance.

From BMW's grill to Pixar's Wall-E

The designs above show acts of visual anthropomorphism, where gesture and expression alone convey personality. They create empathy through closure, a projection of the self as explored in Scott McCloud’s classic Understanding Comics. Pareidolia, the brain’s propensity to recognise faces everywhere, is a powerful trick. Even an oval, two dots and a line create an unmistakable expression; with detail we can add further emotional nuance.

Closure: excerpt from Understanding Comics by Scott McCloud

We can also convey personality through message. In the words of Russell Davies, the rise of devices with personality will lead to a surge in “bubbly writing and objects talking to you in the first person”. Here, an Innocent smoothie prudishly asks us to avert our gaze from its most vulnerable area.

Innocent drinks carton with text "Stop looking at my bottom."

But anthropomorphism needn’t be visual. Consider how R2D2 conveys personality through sound alone – his shrieks and bleeps mapping to human expressions of emotion (See Chris Noessel and Nathan Shedroff at dConstruct 2009 [mp3, 43 minutes]). Similarly, IM programs happily announce incoming messages with a rising fanfare and send replies with a descending farewell.

These can be effective ways to communicate personality, but I’ve recently been reflecting about the fuzzier area of expressing personality through behaviour.

According to psychologist Kurt Lewin behaviour is product of the person in question and his environment (check out Lewin’s equation). Our behaviour changes with context. This suggests that we can only form an opinion about someone’s personality through exposure to various scenarios; a single interaction isn’t enough. However once we’ve formed this mental model, we believe it so thoroughly that we become blinded by it, believing that someone’s personality causes their every action – the fundamental attribution error.

Behavioural variance – acting differently according to our environment – is a celebrated part of being human. Anyone who lacks it is boring. Myself, I act quite differently as a Cardiff City fan than as a grandson, since the contexts are very different. At a party you’re expected to drink beer and flirt with girls, not quietly read a library book, if you expect to be invited back.

Dreary technology

This is why I look at modern technology with mixed feelings. As a tool, it’s unsurpassed. But when we engage with it on any human level, it doesn’t respond in kind. Technology has no behavioural variance and very little personality.

Yes, predictability is a key tenet of usability. High-risk systems must respond to input in forseeable ways: an air traffic control system, for instance, needs to be entirely unwavering. But as we’re learning to appreciate the power of play and emotion in our design activities, is there scope for non-critical technology to display behavioural personality?

Mobile devices, for instance, are increasingly a medium of sensory input as well as informational output. We’ll soon carry devices capable of reading our fingerprints, calculating our position and learning our closest social ties by analysing our SMS and email habits. Adding further richness, recent declarative technology encourages users to publish information that designers can use to build emotional responses:

Google map showing current location as Alton Towers theme park

Facebook status showing a user's engagement

So let’s imagine a Twitter client that asks if you really want to send that drunken tweet (maybe you should have read that library book after all). A mobile that loves going on rollercoasters. An MP3 player that longs to play (and listen to?) a new album for once.

Getting personality wrong

Looking, sounding or acting like a human is desirable only if the human is one we like. Some of our early forays have been spectacular failures. For an archetypal example of botched anthropomorphism, look no further than our most hated paperclip.

Designed to save labour and improve UI learnability, Clippy instead came across as smug and invasive. Not only did his brash tone rub many up the wrong way, but he was irritatingly clingy, appearing on simple tasks where users didn’t need or appreciate help.

The despotic HAL illustrates the other extreme of dislikable machine personality. Clarke and Kubrick created a terrifying villain for 2001 simply by highlighting the unflinching rationality of computation. HAL’s cold-bloodedness is the opposite of humanity. Our heroes are irrational, given to senseless acts in the name of compassion. We can all empathise: who hasn’t done something stupid when in the grip of emotion?

Appealing machine personality lies somewhere between the shores of impassivity and fake friendliness. Social psychology research tells us that we like people who share a similar personality to our own, and people who like us (reciprocal liking). Servile flattery isn’t the answer, of course, but through deep user understanding and reliance on our trusty companions trial, error and feedback perhaps designers will uncover a sweet spot.

We may speculate a few guidelines for conveying personality through behaviour (any additions would be welcomed):

• Personality should be easily overwritten. If you need to make an emergency call, your handset must revert to functionality above all else.
• Personality should be secondary to function. Clippy was disproportionate: his personality overruled his potential usefulness. Not only does this reduce usability, but we risk giving users false expectations of a system’s capabilities.
• Personality should be appropriate to the medium. It may be that desktop computers aren’t an ideal platform for behavioural personality; we still regard them largely as tools of business or home organisation. Mobile phones operate in our intimate space and it’s well known that people form emotional connections with their handsets. Could the mobile arena provide sensible starting points for exploration?

This is largely a thought experiment for now, and it’s clear that behavioural anthropomorphism would raise practical questions. How should users tell devices to stop their shenanigans and get on with the task at hand? Do I want my computer, and whatever systems it’s connected to, to know that I spent the night at my girlfriend’s flat? Would a machine object if I do something it doesn’t approve of?

Any attempt to give technology personality will be divisive. Succeed and we make the technological world a slightly more humane place. Fail, and we create an army of Clippies.

Related resources

• Why is that thing beeping? A sound design primer
• Russell Davies “Materialising and dematerialising a web of data” (mp3, 44 minutes)

Special thanks: Rebecca Cottrell
Photo credits: Duncan, estoril

We are relying ever more on technology to help us out. In this article I am discussing how people form mental models of urban environments, and how technology can augment and even replace our wayfinding skills.

This article is an extract from my upcoming talk at EuroIA 09, The Future of Wayfinding.

Mental Models

Faced with any complex system, we form a mental model. Cities are no exception. Our models (known in the wayfinding domain as cognitive maps) combine cues from across our environment. Some cues are implicit, woven into the fabric of our surroundings: urban density, landmarks, or even the flow of traffic. Others are explicitly designed to describe the structure of a city, such as maps, signs and street naming conventions.

As with any designed system, some cities are more learnable than others. Contrast the regular grid, tall landmarks and self-explanatory street names of New York with the organic sprawl of London:

Maps of New York and London at the same scale (from OpenStreetMap.org) demonstrate the difference in structure of the two cities.

New York has information architecture baked in; London does not. Kevin Lynch named this quality ‘legibility’ – an apt term implying, as does its typographic equivalent, a deep relationship with the identity and DNA of a system.

Survey Knowledge

Good cognitive maps make use of survey knowledge, an understanding of the topological structure of an environment. Centuries of designers have built survey knowledge by printing maps. Some are, of course, more successful than others. The famous London Underground map is so long-standing and ubiquitous that it acts as an ersatz cognitive map for many Londoners. Unfortunately, being designed to show connections below ground, it doesn’t correspond well with London’s surface geography. The map, as they say, is not the territory, and ironically the Tube map hinders effective wayfinding, as people take Underground journeys they would be better off walking.

Survey knowledge gleaned from maps is orientation-specific (this is why maps favour the principle of forward-up equivalence – ‘up’ on the map means ‘straight ahead’). However, we learn areas better by exploring them, which gives us survey knowledge that isn’t based any particular direction. This means that cognitive maps are fluid, changing with context and time to form more coherent wholes.

Different wayfinding tasks

• naive, or exhaustive, search – where the user doesn’t know where the destination is (eg. finding a postbox in a city he doesn’t know)
• primed search – where the user knows the destination’s location (eg. driving to his parents’ house)
• exploratory – where there is no set destination (eg. going for a walk)

User experience folk will no doubt notice parallels to digital information retrieval, including the understanding that most wayfinding tasks will mix these modes. For example we may use a primed search to navigate to a shop found on a shopping mall directory, followed by an exhaustive search for the right aisle within the shop. We can also recognise other concepts from the digital world: the concepts of information scent and berrypicking are both entirely relevant to real-world wayfinding.

Complementing survey knowledge is procedural knowledge: the means of getting from A to B, via C. Sometimes this can be sufficient alone. Plan a route in advance or get directions from a passer-by and you may well find your destination, but if the instructions are flawed or there’s a change in conditions (roadworks, for instance), procedural knowledge collapses quickly and you’re left to improvise or retrace your steps.

Knowledge

Screenshot from the Trails iPhone app

Good wayfinding takes survey knowledge, procedural knowledge and also landmark knowledge, an appreciation of the locations of notable points of interest. Building these three platforms has traditionally been the domain of wayfinding designers, architects and town planners, but now the technologists are getting their turn. Online maps and route planning software have revolutionised the wayfinding business, and computer scientists are attempting to standardise the language of geography through systems such as KML.

While it’s important to know the environment, the user must also know where she is. Technology can be a great help here, with GPS today’s crown jewel. However, although it’s tempting to think that this solves the location problem, GPS is only accurate to 3 metres and, being a line-of-sight technology, doesn’t work indoors or in heavily built up areas. We also need another layer of codification and processing to turn longitude and latitude into human vernacular such as “Junction 12 of the M1″ or “tenth floor of the Empire State Building”.

Mobiles are of course well suited to act as the vehicle for GPS and this codification layer, and have been an understandable vehicle for wayfinding technology. The typical limitations of screen size and user context apply, but the advent of GPS and compass technology in mobiles has led to a sudden commercial interest in ‘augmented reality’, already well on its way to becoming the next misappropriated buzzword.

Nearest Tube iPhone app from Acrossair

Beyond mobile devices, wayfinding provides an excellent stepping stone into the world of ubiquitous computing. Unlike many other ubicomp applications, wayfinding is highly task-driven, meaning many of today’s UCD approaches could be relevant. Imagining a world of ambient informatics, we see thousands of potential output devices. Public LCD displays, signage, buildings, even the street beneath our feet can be our canvas.

The ideal system

The ideal wayfinding system dissolves into behaviour. It requires no inputs, and automatically knows our location and destination. Its feedback to us can take the form of subtle visual, audible or tactile cues – highlighting the path ahead on some display, or even providing a gentle tap on the shoulder when we move in the wrong direction. However, it’s not easy for systems to truly anticipate our wayfinding needs. Although early adopters are habitually advertising both location and destination via services such as Dopplr, FireEagle and Latitude, systems aren’t particularly good at inferring intent. We often don’t navigate rationally – we take scenic routes, stop to pick up lunch, or become distracted by window shopping.

Ubiquitous Computing

We can consider some technologies (such as QR codes, RFID and GPS) as bridging points between the digital world and the real world. At these touchpoints, we are effectively designing an API that allows both worlds to interact. The information architecture must be harmonised, and the correspondences should be carefully aligned. As a prosaic example, labelling and signage used in digital systems must correspond to those used in the real world. It’s clear that the roles of the UX designer and wayfinding designer will start to blur.

Bringing wayfinding into the ubicomp domain might also allow the dimension of time to affect our wayfinding choices. Spimes could help us navigate by highlighting the past actions of others. It’s helpful to know that 95% of all previous travellers to the stadium turned left at a particular turning (of course, this is nothing new to the animal kingdom). Collapsing the past into the present also opens up exciting opportunities to reintroduce our favourite digital mechanisms into the real world. Collaborative filtering, recommendations and other ‘wisdom of crowds’ phenomena could mean wayfinding is no longer a solo pursuit. The notion of anthropocentric wayfinding has impact far beyond technology. It causes flashmobs, football riots and even political revolution. It mobilises us as a combined unit and could even be said to demonstrate emergent hive intelligence.

As with any dream of the future, this picture is utopian and perhaps unrealistic. Much of the world doesn’t yet have running water, let alone a broadband mobile network. Even when these systems are in place, the interaction between people, places and technology will inevitably prove both overly rigid and frustratingly sloppy in various contexts. And isn’t there some joy in getting lost in a new city and stumbling across something beautiful? The advertisers certainly think so:

Finally, in the words of McLuhan, “every extension is also an amputation”. It is conceivable that we will find ourselves relying on technology to such an extent that in the event of its inevitable failure we will struggle at even the most basic wayfinding tasks.

There are clearly challenges ahead. However, getting lost and getting found is an inherent part of human life, and therefore wayfinding is well within the domain of future user experience work. With skill and empathy, we can bring a layer of humanity and usefulness to wayfinding technology.

Interested in more? I will be presenting a session on ‘The Future of Wayfinding’ at EuroIA.