ChatBots: Bright but Inexperienced

Learn how to integrate the conversational channels using 5 UX Guidelines.

Almar van der Krogt
Chatbots Life

--

The channel range; from chat bots and (mobile) web to call center and face-to-face

Conversational commerce and chat bots are nothing short of a revolution! At least, according to Chris Messina, who coined the term ‘conversational commerce’ and makes a convincing case for the surge in chat bots in his recent article. However, he also states that:

“Most bots are too immature to justify their hype.” — Chris Messina

I think this is exemplary of the stage the revolution is in and, as such, no different from the previous digital revolutions and their first websites and first mobile apps.

In that sense it’s history repeating: new platforms by old (eg Facebook Messenger, Google Allo) and new players (eg Slack, Kik, Telegram) are emerging and the crowds are moving. All the more reason to pay attention.

For a sense of the new space see Venturebeat’s Bot Landscape.

Bright bot promises

To be clear about what I mean by chat bots, I refer the Matt Schlicht’s definition in ‘The Complete Beginners Guide to Chatbots’:

“A chatbot is a service, powered by rules and sometimes artificial intelligence, that you interact with via a chat interface” — Matt Schlicht

This means you can ask a bot for the weather, the news, a recipe, to schedule a meeting, what (Dutch) train to catch and even style advice. All these things of course can be done via a webpage or an app, but chat bots do offer major new or additional benefits in comparison to the previous web and app technologies:

  • Natural Language Processing; this means no more, or at least less, need for, potentially unintuitive or confusing, menus and buttons. You can ask the bot in written or spoken language to perform a certain task. Even within an ongoing conversation.
  • Automation; you set the bot out on a certain task, it can converse with other people, or bots for that matter, and report back to you at set intervals.
  • Instant Updating; since bots operate on one or more of the existing Instant Messaging (IM) platforms (which are assumed to be installed by the user already, like web browsers for Internet use in the past) there’s no need to download a new chat bot service; a simple link to or search query for a bot suffices. Consequently, changes or updates to that service are instant. Discovery, i.e. making users aware of your particular bot, does remain an issue though.

Discovery, i.e. making users aware of a particular bot, does remain an issue

  • Self Learning / Artificial Intelligence; because the bot is part of a conversation, instead of a single, pre-set action there is a potential for learning. By measuring the success rates of multiple interactions with different users, the questions and responses of a bot can be optimized and improved automatically. The before mentioned Natural Language Processing is actually already a form of Artificial Intelligence. If it’s applied to a simple decision tree however the interaction range of a chat bot remains limited.

Do I think bots are the new apps: YES, in certain cases. However, they will coexist and cooperate for a long time to come.

For an in depth analysis of why Apps will still persevere see Sarah Guo’s ‘5 Reasons Mobile Apps May Still Rule’.

Location, location, location

The age old marketing adagio ‘location, location, location’, is probably the most compelling reason for seriously considering chat bots; the crowds are moving towards Instant Messaging apps and are staying there. And chat bots are the only scalable meaningful way to interact with these crowds.

Chat bots are the only scalable meaningful way to interact with these [Instant Messaging] crowds

Of course, there is not a single leading platform to be on. As with the browser wars and OS wars before, there are multiple big players and notable newcomers fighting to become the messaging platform of choice. The result so far, as is to be expected, is that there are different platforms for different audiences and different geographies.

The platforms are sometimes, but not always, in line with the prevailing social network for a certain audience or region, which may help in the discovery phase.

Another thing to consider is that the technical capabilities of each platform differ, so not all functions are available on all IM platforms. And, as with all software, the upgrades come in at a rapid pace.

So defining a clear and achievable user and business goal before entering the chat bot arena is a prerequisite. Although Esther Crawford is clear in her article: “Every person with a Facebook Page should have a bot”.

The bot/user (in)experience

A major qualifier in successful webpages and apps is and always has been the User Experience (UX). The same applies for chat bots. And, as mentioned in the intro, there’s still a lot of trial-and-error going on.

The main issue currently is the fact that both the bots and users are ‘inexperienced’ and proper design (even within the still limited boundaries) can go a long way in resolving this.

Below are a few guidelines and ideas that I encountered when exploring and contemplating the new landscape of chat bots:

  • Introducing the bot (and having the user remember it’s quirks); even though the chat interface and the optional natural language processing eliminates a number of interface needs, it’s still essential to teach the user what the bot can and cannot do. This can be specific commands or for simpler bots, the order in which to ask for certain things. In general current chat bots are very task specific, which gives a logical context for the options, but providing a proper response to ‘operator’ or ‘help’ should be common practice (this specific point is also argued by Trish Katz in her article).

A proper response to ‘operator’ or ‘help’ should be common practice

  • Timing; because the chat bot is server based and some tasks may require further back-end processing, there’s a potential delay in the response. The user should always know what is happening and you can’t assume everything on the user side processes instantly either (e.g. receiving an attachment, like a recipe card, in the conversation can take longer to load than the follow-up question, which is confusing). Next to that the users response can be delayed for several reasons as well. If the bot is awaiting a response it should respond after the appropriate amount of time with ‘Do you need help?’ or, which would be ideal for commonly used bots, ‘The usual?’. The latter, of course, would require quite some user specific data gathering and artificial intelligence.

Ideally a commonly used chat bot would, after a while, respond with ‘The usual?’ when being invoked.

  • Degree of anthropomorphism (or how much should the bot act like a human); because the user converses with the chat bot in a natural way it’s easy to attribute human traits to it, even if it’s unintended. I think that at all times is should be clear that the user is dealing with a bot, although it can be charming, and even an unique selling point, if it has character. Or as Yisela Alvarez Trentini puts it (with some uncanny warnings):

“In conversational UI’s, personality is the new UX” — Yisela Alvarez Trentini

  • The hand-off; as with Mobile vs. Web and Web vs Call before that, there will always be a need to switch between channels. Especially when, as in this case, the channel is new for the user it should be easy and straight forward to go to familiar terrain or get actual human support. In most cases the chat bot will not be able to perform all the functions that are available via other channels (and for some cases never will be, as argued by Paul Adams in Bots vs humans), so the hand-off to a more appropriate channel should be taken care of, ideally with retaining as much context and interaction history as possible. Tools to manage this are being developed (like Flow.ai and Meya.ai) to handle this in a scalable way, so you don’t have to leave the bright new ‘bot’-kid in charge of the store without backup.

Don’t leave the bright new ‘bot’-kid in charge of the store without backup

  • Graceful degradation (or keeping your cool when things go wrong); as you will notice when playing around with chat bots, their limitations are still fairly obvious and especially the natural language processing is still in the early stages. And what you don’t want to happen is the user is getting stuck with an seemingly ignorant and hard hearing robot, stating “I don’t know what you mean by…..” or “Sorry, I didn’t get that” (for more examples and cases see Pep AI’s article). Re-iterating the basic commands of the chat bot to the user is a first option, but especially in these early days learning should be central to the interactions of the chat bot. Next to the potential machine learning or artificial intelligence in the background there is an opportunity for ‘traditional’ learning. By simple asking the user ‘What answer or action were you expecting?’ (as you would in a poll or survey) after an interaction seems to have gotten stuck, can give many insights for improvement and gives the user satisfying ending to the conversation. Or you do it the Star Wars way:

C-3PO: I beg your pardon, but what do you mean, “naked?”

R2D2: [beeps]

C-3PO: My parts are showing? Oh, my goodness, oh!

The future: Her vs. HAL?

I think there’s an exciting future ahead of us when it comes to chat bots and eventually full blown artificial intelligence. I’m not sure how much truth there is in the dystopian examples Samantha from the movie ‘Her’ and HAL9000 from the movie ‘2001: A Space Odyssey’ (who lives on on Twitter 😜), but this could happen if the two would meet:

--

--

I’m living la vie virtuelle @VIRVIE. Passionate about #netart and venturing into #bots #AI. link to my 🤖 alter ego https://m.me/almarbot1st