A chatbot state of mind

Giving thought to what goes into designing an interaction on any visual interface… until recently, it was about defining the type of interaction (business function), goal (business outcome) of the interaction, default/preferred medium or channel (that the business uses) and the workflow/s (business SOP) to adapt.

UX and design thinking had finally arrived. Every recognisable business highlighted their focus on user experience, design thinking, process awareness etc. So then, how is it that with the advent of AI, Chatbots and the mad rush to create automation that mimics or supports humans, we have missed out on the user?

Don’t get me wrong- we have identified and are looking at the user, but still miss out what s/he is thinking, feeling, over and above what has been identified as user ‘need’.

Chatbot builder tools and platforms enable developers and designers to follow the standard process of building a technology solution; Identify an opportunity, and adapt a way of delivering a solution based on our (the businesses) expectation of a resolution — a ‘technically enabled’ automation of an existing process. The result? We have chatbots churning out real-time insights on large amounts of data processed, to support call center agents make better FCR (first call resolution) figures, optimise escalation handling, and dealing with contextual conversations. However, its interesting to see if these chatbots/AI systems engage with the consumers/endusers in a way that ensures they stick with the business? Are businesses leveraging technology to truly connect with their end users? Are these automations making the consumer-enterprise conversations better?

Companies like a Facebook have pioneered and mastered that art of processing data from your timeline to show you product recommendations via their advertising partners. They’ve also however messed up by sending the same users inappropriate ads based on a pre determined lifecycle of the said event insight, when they sent one user baby care product recommendations even after the user had shared an unfortunate event about a miscarriage a few weeks earlier on her Facebook timeline. The issue was, the algorithms while being able to pick triggers from the initial data, was not built to consider future data on the same users timeline to contextualise the possible scenarios on that particular event lifecycle. A design thinking exercise would have surely helped here, as you would be able to map out various user specific scenarios and emotions to deal with across any life event.

Not to digress further, the point I’m hoping to make is this- chatbots or any automation of processes needs to be planned out from a user experience methodology that considers scenario probabilities and user emotion at every point in time. I call it a ‘state of mind’ analysis of the user, so that UX designers can help conversation designers to adapt the chatbot behaviour and narrative to help engage users with empathy. Going back to the call centre scenario, for an AI to assist an agent to engage with the customer using relevant recommendations, suggestions, etc. the system needs to be designed to relate each conversation to a real-world human interaction, and maybe then, we would see the AI assistant recommend ‘empathy guides’ on screen for the agent to work with.

A basic methodology I’ve followed to help design our chatbot conversations involve considering three typical possibilities to any user conversation —

A. What is the user asking for/looking for?

This can be derived by basic intent identification, where the bot knows of the type of information the user needs

B. If the user is asking a question specific to him/her, the bot needs to check and see if there are any possible actions that can be recommended to the user, depending on the users statement.

C. If the user has requested an action, the bot should be able to reference possible results of that action, as well as check for business rules to recommend or ask the user for further data so as to help with that trans/action.

Now when we overlay the UX component for each of the above scenarios, one can apply it to any specific industry to define a state of mind driven conversation.

For eg.- In the case of a person chatting with a banking bot and reporting a fraudulent transaction on his/her card, the base assumption of the state of mind of the user is that of ‘concern/distress/anxiety’. The bot needs to be able to handle the conversation with utmost care so as to quickly record the issue and also assure the customer that the issue has been noted. Initial Reporting can be handled with minimum validation (just a mobile number or valid email id should suffice) . The bot should then ideally communicate to the user, the process that will be followed by the business — this ensures that the user gets an idea of the timelines for the resolution process and updates. This type of a conversation would never be closed with a ‘rate us’ survey related conversation in the same or next session (unless the issue is resolved by then).

Scenario 2- the user is looking for a loan and the bot it engaging him/her in a conversation. In this case, the bot should ideally be enquiring with the user about their specific needs, and if the data on their profile is available with the bank (eg. Latest credit / eligibility score), then the bot should be able to instantly recommend options without having to dive into a data collection flow. In this case too, the users state of mind would be one of ‘exploring/ caution / tentative ‘ as it is a new product, a commitment, and a liability s/he is taking on. The bot conversation needs to be assuring and encouraging while leading the user towards a decision.

My favourite is the scenario where the user is asking for something and the business doesn’t have that, but still engages in a conversation after the initial ‘no/sorry we don’t have that’, gives some recommendations based on the users history, profile, or current conversation to lead the user on to other possibilities. Conversations like these are mimicking typical sales chit chat- where one goes in looking for an D’addario guitar string pack and comes out of the store with an Elixir string pack plus a new tuner and a clip on amp connector. Every buying decision we take is usually an emotion or ‘state of mind’ based one. Every experience we remember is also completely a state of mind related one, good or bad. For a chatbot to do the same, the domain expertise, a real time understanding of the user and their state of mind, becomes key in order to deliver user delight.

For conversational AI to truly reach its potential, our responsibility as designers and bot creators would be to dive deep into the users mind, and by relation, the chatbots state of mind — be the user to build for the user.

Related posts:

Vortexes are spiral forms that swirl around an invisible center line. You can find them in your bathtub when water drains and even in cloud formations above Saturn’s north pole. The vortex in this…

Letter of redemption from an ex-husband.

Hearing the way he conducted his cold call made me reflect on the way I do things. Parts of it were great, and parts I thought could have been better. The first thing he said after his initial…