Evolution of ARM: Up Next, Collaborative Intelligence

The way we use words and phrases changes as our knowledge about the world around us evolves. For example, the first use of the term “gluten-free” appeared in 1927. Over the next 80 years, numerous studies about celiac disease and non-celiac gluten sensitivity emerged. And in 2013, almost 90 years after the term was first used, the FDA finally issued rules for labeling gluten-free food items.   

Such is the trajectory of language: Words and phrases enter our linguistic periphery before they enter our lexicons with their usage finally solidified and agreed upon — at least for a moment. 

The term “collaborative intelligence” is no exception. The concept originated in 1959, with Oliver Selfridge’s famous Pandemonium: A Paradigm for Learning, but the term itself only became more widely used and accepted following the coining of another, related term in 1994: collective intelligence. 

Pierre Lévy, who coined the term "collective intelligence," proposed that collective intelligence encompasses both collaborative and collective intelligence. The two are different, and for the purposes of this article, it’s important to create clear distinction: 

Collaborative intelligence refers to distributed systems where all agents contribute to a problem-solving network (autonomously or not), while the knowledge produced by this network can be referred to as collective intelligence. 

Think of it this way: If collaborative intelligence is the hive mind, then collective intelligence is the resulting knowledge this mind agrees upon and applies. 

And why does it matter?

As awareness of the term “collaborative intelligence” continues to push its way past the business periphery, the ideas it involves will make similar usage gains (and vice versa). And in an industry soaked with “artificial vs. human intelligence” rhetoric — and the fear and immobility that rhetoric perpetuates — we need to commit to an ideal of collaborative intelligence now. 

Doing so can pull us away from debate and toward ARM intelligence that makes a real difference in the way businesses function and communicate with borrowers, debtors, patients, our companies, our data, and yes, of course, our NLP and other ML models. And that way of functioning will also change our results. 

What does committing to collaborative intelligence look like, in practice? 

Collaborative intelligence requires you to:

1) Believe in the inherent value of diverse information, and

2) Redefine transparency

The Inherent Value of Information Diversity

It’s a fact: Solving complex problems demands individual expertise, the incorporation of conflicting stakeholder priorities, and the differing interpretations of experts with diverse lived experiences. In a system where that set of perspectives includes an AI model and an automation workflow, the trust you place in that diversity is paramount. You cannot reach collective intelligence and support positive ARM business outcomes without widening your lens. 

A Redefinition of Information Transparency

Transparency is a tricky topic in ARM. Who should see what data, and when? While a collections agent shouldn’t expose consumer credit information to their social media network, for instance, your notes should expose data that will reshape workflows to your workflow automation model — and then expose the results of that automation to your human agents for their own application.

How do the agents in the system (your compliance department, their QA team, your contact center agents, the robotic process automation that augments those agents, your workflows, etc.) know when and where to draw the lines? 

Transparency must be redefined. For the ARM industry to continue its evolution past simple “human vs. robot” conversation, we must allow transparency to act as a driver for business outcomes. What will you achieve when all agents in the network are empowered to both contribute and use the right information, at the right time? 

You (or your model) might be able to create a connection between a set of conversations with a borrower and a future QA review workflow. Or a borrower may receive the exact right, personal response for their emotional state through exposure of their contextual data to a human agent. By treating transparency as a driver, you can decide what to correctly expose and when, and give all agents (human or not) the power to do the same.  

Collaborative Intelligence is Here. Are You?

Here’s the bottom line: A commitment to developing and working within true collaborative intelligence systems will improve ARM. 

Consider the virtuous cycle that will occur when all agents within the problem-solving network are enabled with information. That information improves workflows for productivity, training, compliance, and so much more. And those improvements inform person-to-person conversations — the data from which informs improvements to workflows. 

This virtuous cycle both originates from and becomes the system itself — a system that produces the kind of collective intelligence the ARM industry requires to thrive.