IA Summit: Creating the Adaptive Interface

by Luke Wroblewski March 23, 2007

Stephen Anderson discussed several possibilities for (Micro) Adaptive Interfaces during his The conversation gets interesting: Creating the adaptive interface at the 2007 IA Summit.

  • Most applications choose to reside in the zone of safe mediocrity: they work ok for most people. But what’s convenient for some people might not be for others. What’s common use for some might not be for others.
  • Micro-adaptive interfaces do not change applications drastically, instead they adapt to long term behavior in subtle but important (useful) ways.
  • Example: if we know users frequently attach an average of 2 files to their messages, we can default them to two upload fields.
  • Example: If we know people looking for information from a particular context (all information once vs. find a specific value in a table), we can adapt the presentation accordingly.
  • It’s not that difficult to create personalized user interfaces: If (this happens/we know that) then. But we need to have the data to make intelligent adaptations.
  • Some ideas for adaptive interfaces:
    • Text box that expands as you type into field: multi-line, or single form field.
    • Default country-selection based on IP address: pick the right country
    • Lists that dynamically change based on user behavior
    • If a drop down gets too long, could switch into auto complete.
    • Slider widget to set date range: could hide accessibility features for people who don’t use them
    • If notice people are missing a button, we could make it bigger.
    • BBC paths changes color based on use. Color and saturation to indicate age and importance of documents.
    • Change prominence of help links: fade out over time if not used.
    • Interface changes based on point in time: confirmation stage of travel reservation vs. navigation stage (getting to location)
    • Shorten driving directions based on frequency of use
    • Can we change help text based on type of users? Professional tone vs. engaging or friendly.
    • Change labels based on location. Example: coke, soda, or pop
  • Given the right data, all of this is relatively easy. Can use cookies, profiles within a site, Mac Address, IP address, marketing data, and internally collected data & usage patterns.
  • Data seepage: multiple bits of small behaviors can lead to more complete data understanding.
  • Why now? More bandwidth/storage, more applications built for specific audiences, open source (can re-use and build on), cleaner separation of content & presentation.
  • Get the basics right first before trying this stuff. Micro-adaptive interfaces are more like a fine-grained sandpaper.
  • Communication & disclosure: let people know what is going on.
  • Doing it wrong is worse then not doing it all.
  • Be careful how you interpret data.
  • Using a better widget in one place might make another part of the interface crummy: think about context.
  • Consider customer support implications: How to support people when the interface changes?
  • Keep it small. No big changes.
  • Be wary of messing with spatial organization: moving things around confuses people.
  • When dealing with data don’t reveal too much of what you know too fast.
  • Check out Identity 2.0, Game Design, Analytics/data mining.