Automating a Culture of Continuous Feedback
How automating feedback with AI powered conversations can aid decision making in real-time
I’ll be speaking about AI, feedback & regulations at The Next:Economy Summit on Oct. 10–11 in San Francisco. Come be part of the discussion about technology and the future of work.
All systems need feedback to learn, improve and course correct. The autopilot functionality in driverless cars is a perfect example. Sensors measure the desired speed and position of the vehicle — among other indicators — and send that data to control systems which adjust accordingly.
Feedback in systems that involve people gets messier. Consider this classic example of city scale feedback from civic tech: drivers see potholes and tag a map with the location, which signals work crews to come fix the issue. If only all our goals were as clear and uncontroversial as “safe roads”. If only the path to meeting that goal were always as clear as a bit of asphalt fill. If only all our problems had a geolocation.
But, alas, the goals of an organization, enterprise or government may be ill-defined or change. Vision and direction may be vague or inconsistent. The lag between signals, decisions and downstream effects can be months or years, with thousands of intervening variables. People have competing interests and different interpretations of events.
Gathering rich, organic feedback on a continuous basis is necessary for managers and regulators to make informed decisions. Robust feedback means honoring people’s authentic voices, rather than shoehorning them into a multiple choice format. It means taking the time to find out how many others share what may be a surprising opinion (to management) or understanding of a situation. It means preserving minority opinions. It means listening well.
But getting rich feedback from a population usually starts with in-depth interviews of a representative sample. Surveys are then created based on the interviews to see which ideas are representative. The process is manual, time consuming and requires specialized knowledge — in a word, expensive.
Applying artificial intelligence to the problem of gathering insights from large populations alleviates much of the burden. AI assisted feedback gathering means no routine effort or process on the part of the organization. It realizes, in yet another domain, Licklider’s vision of human computer symbiosis:
Licklider[’s] … vision was to enable man and machine to cooperate in making decisions, controlling complex situations without the inflexible dependence on predetermined programs. … Licklider foresaw computers doing all the routinizable work that was required to prepare the way for insights and decision making. — Shyam Sankar, TEDGlobal 2012
It means genuine, rich feedback can appear in the inboxes of those who need it, from those who are best positioned to give it, every day or every week — as often as makes sense. It also means that feedback can be scripted or scheduled to run in a certain way at a certain time, in extremely flexible ways.
Here’s an example of PolisBot for Slack, an AI facilitated Slack conversation, outputting a synthesis of user activity, produced from thousands of user interactions:
pol.is is designed from the ground up to be a technology that can completely automate getting feedback from groups of people of any size, even tens of thousands. Humans ask questions that come naturally of large populations, and receive digestible results that they can use to ask further questions.
What does this look like for participants?
Participants ‘agree’, ‘disagree’ or ‘pass’ in response statements made by others. Each participant can also choose to submits statements that represent their viewpoints. That’s all there is to it. Machine intelligence handles the complex work of synthesizing everyone’s views into something digestible.
Conversations can last as long as necessary, and people can take part whenever they want.
Imagine a regulatory agency wants to engage thousands of on-the-ground experts in a weekly conversation about topical issues, to expand its knowledge network. pol.is is currently built for web and Slack, but using the pol.is API, the following custom experience could be created for SMS. The following conversation could be texted to a list of known participants, for example teachers:
What can automatic feedback tell us?
First, it means that feedback can be transformed from something organizations have to ‘do’, ‘develop process around’ and ‘prioritize’ to being part of an organization’s environment. Further — because conversations can be scripted and scheduled — it’s possible for feedback to simply occur at an agreed upon point in the future.
“Siri, could you check in with the company in three months re: the decision to switch insurance providers?” — now a few API calls away
This type of automated feedback means that deep insights synthesized from hundreds or thousands of people on the ground can regularly show up in the inboxes of people high up the chain, allowing for continuous feedback and, hopefully, continuous improvement.
An airline or hospital network could ask thousands of pilots or nurses about a policy change and how it would affect their ability to work while it was still just an idea. Better real-time information about possible changes prevents wasted efforts and unintended consequences.
It means that organizations can capture dissent, refocus assumptions and break apart group think as a part of their daily routine. A high resolution portrait of what’s going on in an organization can now be made automatic.
Most importantly, though, it means organizations have the potential to learn and improve on ideas more quickly.
Happy automating! Follow us on Twitter: @UsePolis.
