3 Ways to Increase Collaboration with AI

A recent study conducted at the MIT Lincoln Laboratory found that humans disliked playing Hanabi, a cooperative card game, with AI. Instead, the participants preferred collaborating with a rule-based bot. The score at the end of the games was not statistically different between the two types of partners (AI vs rule-based agent). But the human participants, all of whom were skilled at Hanabi, “found [their AI teammate] to be unpredictable, unreliable, and untrustworthy, and felt negatively even when the team scored well.”

This study highlights some of the issues employers and programmers may face when implementing AI in the workforce. AI, including augmented intelligence, has immense potential to change how we work and play. As computer scientist Andrew Ng famously said, “AI is the new electricity.” Unlocking this potential requires collaboration between man and machine.

How? Let’s look at three ways organizations can implement AI—without negatively impacting company culture or human workers’ attitudes.

Explainable AI

One of the biggest challenges of AI is the inability to understand why the algorithm arrived at a specific decision, which is referred to as the black box problem. Ensuring humans understand why AI made the decision it did may help human trust their AI coworkers. AI with the ability to communicate its decision-making process is called explainable AI.

Updated Processes

Plugging AI into current, human-only processes might work, but it does not take advantage of AI’s full potential. Instead, introducing AI into a process should prompt a deep look at how the process can be updated and adapted to the unique strengths of AI and humans. When Mercedes-Benz introduced AI-enabled “cobots” into its manufacturing processes, the company shifted the role of human workers to become guides, while the robotic arms take care of moving heavy parts.

The Right Problems

In order to collaborate effectively with AI, humans must first decide and define what problem they want to solve with AI. The most pressing problem is not always the most obvious. For example, a large agricultural company wanted to use AI to predict future crop yields. This is a task AI is more than capable of. However, it probably would have left farmers frustrated, because their most pressing problem was actually how to increase productivity.