The first step towards building contextually responsive digital experiences is to establish a baseline of quality achievable even when there is no contextual information to respond to. As paradoxical as this may sound, the very nature of mobile devices and wireless networks is such that there will always be times when the ideal data to drive contextual adaptation are unavailable. Digital experiences which fail in these circumstances lose the positive sentiment earned while contextually intelligent and risk permanently alienating users.

Just as the principles of ‘mobile first’ educated designers to focus efforts on the quality of visual experience available to the most resource-constrained devices, so too should designers experimenting with contextual response consider what can achieved with minimal or no contextual inputs.

Contextual guesses enhance an experience, but they are no substitute for basic quality considerations. Therefore a principle of progressive contextual enhancement, built on a base quality of standard experience, should be at the heart of design strategy.

The importance of this principle was highlighted to me by Sabrina Majeed, designer at social TV start-up Miso and a speaker at the September 2012 MEX. Miso’s first iterations concentrated on accurately identifying the users’ exact TV watching context, down to matching specific interactions on their mobile device to a precisely timed moment in the TV content.

However, subsequent iterations have seen Miso moving away from relying purely on responding to this ‘ideal scenario’ and focused attention on ensuring the app always offers a useful experience, no matter how few contextual inputs are available. Miso’s design strategy is built around enhancing the experience through context in a modular way, but without the risk of permanent failure if the app has to operate without live contextual data.

Organising contextual inputs into categories based on their relative availability helps refine design strategy.

For instance, data derived from local sensors where an absolute result is certain, such as screen orientation or battery level, are less likely to fail or be misinterpretted than, say, data which are more variable or dependent on external factors. A device may be able to tell an app it is connected to a Wifi network, but be unable to indicate whether the internet connection is working properly (e.g. if connected to a public hotspot, but not yet authenticated).

Designers, just like historians and journalists, should question the source of their contextual data and build services which respond accordingly.

In the example above, it would be safe to make major aspects of the design reliant upon knowledge of screen orientation, but risky to do so in response to knowledge of Wifi connectivity. A prudent designer would, therefore, avoid tying key features of the application to this contextual scenario without a strong fallback position.

A similar principle could be applied to differentiate contextual data specific to individuals and those obtained in aggregate. As an example, a search application may know that the vast majority of customers in a railway station, whose trajectory followed a train line for the 30 minutes preceding an internet search query, are looking for a taxi. However, there are numerous places in the UK where cycle paths run parallel with train lines and a cyclist initiating a search query from a station will be looking for anything but a taxi.

In this scenario, designers would ideally build the architecture to allow individual context to override the contextual average (e.g. an on-board motion sensor could detect whether the user was cycling or traveling by train). If such data could not be obtained, the application could revert to generic search, well adapted to any individual, rather than pushing an unsuited user into a search category that was likely – but not certain – to meet their needs.

Remember there are occasions when users will prefer to remain contextually anonymous rather than be mistaken for someone they are not.

Another consideration which has emerged while working with Qualcomm Research Cambridge is the importance of how contextual data are obtained. Currently, contextual sensing is associated with high power consumption and intensive processor use. This is primarily because today’s contextually aware experiences rely on data sourced at a relatively high level within the software stack.

As a result, making an app contextually responsive has hitherto required compromises in other areas of user experience. Designers have, rightly, been reluctant to make these trade-offs. Users whose device runs out of battery half way through the day are unlikely to look kindly on even the most intelligent contextually responsive experiences if they associate them with power drain or slow performance.

However, Qualcomm’s background in processor design is enabling it to optimise contextual sensing at a much lower level within device architecture. It is an example of how companies throughout the mobile value chain, right down to chipset level, have a critical role in shaping the overall user experience. This broad engagement with the value chain has been at the heart of the MEX initiative from the outset.

Another perspective on this topic emerged during conversations with Louisa Heinrich, Group Strategy Director at design agency Fjord, and also a speaker at the September 2012 MEX. Heinrich questioned how users will feel emotionally about contextually responsive experiences and the importance of design in maintaining a human quality within even the most data-driven systems.

She prompted me to think about how, even when contextual responsive design achieves functional excellence, there is a further responsibility to shape the emotional quality of the user experience. Helping users to understand how a complex web of contextual data are influencing their digital interactions is a multi-faceted challenge of creating trust, implicit orientation of their position within a system architecture and the overall sense that they are masters, not slaves to contextual response.

The exploration of these issues continues in MEX Pathway #14, entitled ‘Define techniques for context aware user experience‘. There will be further essays and links in the MEX newsletter, which you can follow on the blog, by email or on Twitter.

The focal point comes at the next MEX on 19th – 20th September 2012, where speakers including Colm Healy of Qualcomm, Louisa Heinrich of Fjord and Sabrina Majeed of Miso respond to this theme. Working sessions, led by Peter Andic, provide an additional dimension of in-depth problem solving, where a team collaborate on creating a specific response to the Pathway challenge. Early bird tickets, priced at £999 and saving £500 are available until 21st August.

I’d love to hear from anyone with thoughts on this area, so please get in touch to share ideas. Here are some of the initial Pathway provocations to get you thinking:

1. In which scenarios do real-time sensor data and behavioural patterns combine most effectively to reveal user context?
2. What is the right balance between dynamic, contextually responsive elements and consistency in the user interface?
3. How can contextually responsive experiences respect privacy yet learn user behaviour quickly enough to deliver benefit?
4. How do graceful failure techniques allow users to correct inaccurate contextual guesses and reassure the system is learning from mistakes?

(This essays builds on my original article, ‘Context‘, published in April 2012.)