Transforming User Experience with Chatbots
Natural language processing (NLP) is entering the mainstream and the virtual assistants and chatbots which NLP empowers will eventually change the way we seek information from the net by steering us away from traditional apps - many of which are used more and more sparingly.
It’s becoming increasingly apparent that more fluid ways of transacting with our phones such as voice control could offer new options for internet marketing.
Gartner predict that by 2019, 20% of brands will have abandoned their mobile apps due to low returns on investment; a claim supported by a recent Microsoft study which held that 85% of smartphone usage is channeled through just five regular apps.
Rather than compete with these ‘top 5’ apps, which are often messenger type services, Microsoft suggest offering a service such as a chatbot on top of a service such as Skype, Facebook Messenger (over 1 billion users), Apple’s iMessage, or Whatsapp.
Apps are clearly an essential way of reaching customers in industries such as media, gambling and banking, but for companies which rely on websites and banner advertising, chatbots could become a highly favoured option.
The abundance of new chatbots appearing on Facebook Messenger promote the idea that your bots could eventually exist alongside human contacts.
How it Works
Right now, Facebook’s WIT.ai, Micrososft’s Luis.ai, and API.ai are amongst the main technological frameworks from which those curious about the technology can explore its capabilities and build functioning products.
These development frameworks use combinations of Natural Language Processing (NLP) engines which enable the user to set out common parameters such as ‘intents’ (what the user is trying to do) and ‘entities’ (specific information needed to understand the request). This backend technology will be trained and informed by middleware which provides the ability for conversations rather than simple challenge/response requests.
The end result will be a chatbot which is hard to distinguish from a genuine human correspondent (as long as your query is relevant to the applications subject matter!).
Platforms like Facebook and Microsoft favour mass production which means that the underlying tech is built using fairly generic domains which are hard to adapt to specialist knowledge, and the big offerings are inevitably closed source. Nevertheless, they do offer the chance to acquaint with the technology and they provide a stage for highly comparable products.
How Chatbots Could Transform User Experience
Most of us have had experience of verbal communication with technology- for instance when verbally asking Siri, Alexa or Google about the weather or cinema listings - gleaning answers which require far less ‘point and click’ style navigation through search results and websites.
From a UX perspective, chatbots could have huge benefits for those with low IT literacy with verbal communications much more suitable for those without that ingrained sense of how to navigate the net.
The ways that humans are required to interact with technology can be vastly different from site to site. If language effectively becomes the interface, then the inputs derived from the user could be much more unified and consistent.
Gartner recognise the imminence of language as an interface, predicting that by 2020, 30% of browsing will be screenless.
Orange Bus UX Designer, Kevin Webster, made the following observation about the potential for UX to contribute to Natural Language Processing:
Voice commands, in general are growing in popularity: GoPro’s new devices - the Hero 5’s - use voice controls for all of their basic operations, and Xbox has popularised the technology by integrating voice commands into its media selection and functional Operations.
The efficiencies this kind of technology can create will benefit high volume users - through rapid scheduling and instant access to information, but one of the biggest challenges will be ‘can i trust this?’. It would only take one appointment or note saved incorrectly for this technology to fall out of favour.
The technology still follows a system design approach at its core and relies on machine learning for self-improvement, but there’s a big responsibility from a UX perspective - helping to understand the types of conversations users might want, along with robust journey mapping of actual user behaviour.
Providing responses acceptable to the system’s understanding comes with it’s own challenges. On a website if you forget to add some information or your input is incorrect you are notified and directed to the area of field to be corrected. However in a spoken conversation this can break the narrative and can feel out of place, even awkward. You have no context to the other questions you may have answered increasing cognitive load on the user to recall the answers by memory. The illusion only works if it feels natural.
The linear paths (mental models) we’re used to are more woolly with chatbots and voice control. For example, ordering a taxi is simple with Uber but opening that up to a conversation could make the input unpredictable and prone to mistakes - we often rely on the software’s smart defaults to fill in the gaps in our own knowledge. Machine learning and error prevention will still require careful monitoring.
Consistent user testing will also be important to the development of this technology. It’s more than likely you’ve tried to break or trick Siri or Cortana, reflecting our human nature to challenge and find the limits of what we use. In my opinion, the extra considerations and work will be worth it removing other issues that can arise from information architecture, usability and content.
Accessibility to this technology is easy for UXers as you’ll have the audio and the transcripts from which to base assertions.
Messaging apps are a massive internet growth area and this is clearly where the initial battleground for chatbots will lie.
Right now, the best Facebook chatbots tend to revolve around simple services such as news feeds, weather enquiries (Poncho) and personal assistants (Ozlo), but the leap to the likes of public frontline services may not be too far away.
The failure of NHS Direct’s web based symptoms advisor in 2014 was an example of an attempt to use context aware technology before the right NLP type solutions had matured - now that chatbots are increasingly sophisticated, services like this can be reborn.
A future where chatbots are actively seeking to communicate with users is on the horizon, but for now natural language processing can begin to meet the public in the middle. Orange Bus strive to be at the forefront of these developments and are already pioneering solutions where we believe that chatbots may be a perfect fit.
If you would like to get in touch with our UX team or have a project which is seeking to explore new digital technologies, contact us on 0191 241 3703, or email firstname.lastname@example.org.