The Future of CUI: How AI Will Evolve Beyond Chatting to Actual Dialogue

We’ve all seen them in sci-fi movies: C3PO from Star Wars, Hal from 2001: A Space Odyssey, Rosie from The Jetsons… These characters represent the epitome of artificial intelligence (AI), not just because they speak but because they converse.

Chatbots and other forms of talking digital entities have been around for a while, but only recently has the technology evolved to allow AI to engage in actual dialog. In fiction, this is great for developing engaging characters. However, in the real world, it represents a paradigm shift in the user interface. No longer will we need specialized skills to tap the power of the digital ecosystem. 

If you want something done, you don’t need to code or even navigate through complex systems and menus – you just say it. Not only does this come in handy in life-or-death situations (“Shut down all the garbage mashers on the detention level!”), it serves to streamline numerous mundane functions in our personal and professional lives (“How many widgets did we sell in the Pacific Rim last quarter?”)

Enter the CUI

In technical parlance, CUI refers to Conversational User Interface. By combining large language models (LLMs), natural language processing (NLPs), and emerging fields like natural language understanding (NLU), we have reached a point where AI can do more than just respond to queries. It can engage in discussions and even lead users to conclusions and courses of action that they may not have considered before.

According to ZenDesk, the CUI offers tremendous potential across a wide range of enterprise applications. Not only does this provide a more seamless and positive user experience, but it can also be coupled with things like voice recognition and biometric authentication for identity verification and other security measures. The CUI can as well be leveraged for enhanced data collection and analysis – divining not just customer histories and patterns but real-time interactions and behaviors – all while maintaining a friendly, accessible, and unflappable posture.

Like most forms of AI, however, don’t expect this level of functionality right out of the box. CUIs must be trained extensively in order to develop a consistent and coherent conversational flow. From there, it must draw off of large data sets to provide the kind of personalized and context-aware service that meets or exceeds that of a human interlocutor. And, of course, this data must be accurate and up-to-date.

Still, Eric Griffing, growth marketing manager at conversational AI platform developer Dashbot, says that true conversation relies on more than just raw data. Normal bots mostly leverage machine learning, which is fine for analyzing large data sets and absorbing conversational information in real time, but without a rules-based approach to the interaction, they can get tripped up fairly easily. 

By applying a properly defined set of rules to guide responses, chatbots can make the leap from basic chatting to rapid determination of the kind of use it is dealing with (a customer, provider, agent) and tailor its responses appropriately. In essence, it uses the same kind of intuitive rules that people apply – often unwittingly – when they hold conversations with each other.

Griffing added: 

A Matter of Design

Even with the right technology, however, CUIs must be designed properly. Bo-Peter Laanen of the Conversation Design Institute argues that a key factor in good design is the concept of inclusivity. Inclusive language models incorporate not just multiple languages but idioms, slang, and all the other vagaries that inhabit human speech (and thought). 

With this in hand, the bot has the ability to communicate in any way that resonates most clearly with its user, and a single platform can be distributed across wide-scale environments with little to no retraining or recoding. In fact, it can act as a translator across language and generational barriers.

This could very well be the secret sauce that turns conversational AI from a novelty to a ubiquitous tool, says author and researcher Natalia Borysko. In surveys carried out for Eleken Designers, the two biggest detriments to the widespread deployment of current chatbot technologies are: 

• The cost and effort to explain issues in ways the bot can understand;
• The high rate of failure to resolve issues successfully. 

To that end, Borysko has devised nine ways to make CUIs more “human”:

• Make them more specific: By tailoring bots to specific fields, they can more adequately address the needs of users interested in a particular topic or function.
• Create a multiple-choice environment: Open-ended conversations often lead to dead ends. Guiding users through multiple options is the quickest and easiest way to produce successful outcomes.
• Add emotions: This will make responses less dry and more engaging.
• Enable feedback and rewards: If a bot can identify peak moments in a dialogue, it can then encourage or congratulate the user for a job well done.
• Give a helping hand: A bot that can offer suggestions for workarounds and all-new courses of action will come across as friendlier and more useful.
• Standardize vocabulary: Consistency in terminology will allow users to more fully comprehend what they are expected to do.
• Take the right tone: Some voices and personalities work better for certain tasks than others. You don’t want to answer questions about medical issues or finances in the same voice that provides movie times.
• Add personality: The more unique your CUI is, the more welcoming it becomes for both new and repeat users. It also goes a long way toward defining your brand.
• Admit defeat: Establish a way for the bot to disengage gracefully when it cannot meet the user’s request. This includes redirecting him/her to another resource.

The Bottom Line

The ultimate goal of the CUI is to smooth over the rough edges in human-digital interactions. By doing this, we should be able to democratize computing power on a scale that dwarfs the introduction of the PC and the Internet. Even young children will be able to query complex data sets and perform intricate tasks with only the simplest commands – within the right guidelines, of course.

And since the digital world will soon inhabit virtually everything around us – from our automobiles to shoes – we all should have little problem getting what we want without having to beg, plead or argue.