Steve King
Group Intelligence for Human + AI Collaboration
Steve Jobs designed the office structures and atrium at Pixar so it forces people with different skills and viewpoints to rub elbows and socialize in ad hoc encounters.
With the help of a great workspace, Pixar executives developed a meeting culture that mandates peer-to-peer exchanges. Everyone in the room contributes, without fear of existential threat, before the meeting is adjourned.
After Pixar and Disney merged, Pixar exec Ed Catmull said it took Pixar about 4 hours to explain their collaboration framework and about 4 years to actually deploy it in the Disney context.
If collaboration is this hard for the top players... what about the rest of us!? Does the introduction of AI into knowledge workflows make cross-discipline mind-melds harder or easier?
Collaboration design... not trivial stuff! And now we have to collaborate with AI in unfathomable ways.
Managing Complexity with Human/AI Collaboration
Ever growing complexity in today's infrastructure and organizations calls for a new set of human/AI collaboration approaches. Group Intelligence software enables team platforms that scale human and AI expertise, moving cross-discipline knowledge-bound workgroups through structured brainstorm exercises with the goal of shared problem solving, co-creation and peer-to-peer decision making.
Group Intelligence software can be more effective than Slack, Yammer, Jive and other social collaboration tools, which are essentially just chat and content sharing... not facilitated human/AI problem solving.
One of the key principles of Group Intelligence is the pattern of converging and diverging ideas created by the group.
Cognitive patterns of Human/AI collaboration
Expanding the scope of discussion is known as "divergence" in facilitator-speak. Narrowing down the concepts for discussion is "convergence" or "reduction" ...key principles for human and AI inference activities.
To navigate a shared reasoning experience, team members must jointly engage in a sequence of cognitive thinking patterns that are supported by neural structures managed by the prefrontal cortex (Damasio and Damasio, 1992; Holyoak and Kroger, 1995).
Diverge — move from having fewer concepts to having more concepts.
Converge — move from having many concepts to focusing on a few concepts deemed worthy of further attention.
Organize — move from less understanding to more understanding of the relationships among concepts.
Elaborate — move from having concepts expressed in less detail to having concepts expressed in more detail.
Abstract — move from having concepts expressed in more detail to having concepts expressed in less detail.
Evaluate — move from less understanding of the value of concepts for achieving a goal to more understanding of the value of concepts for achieving a goal.
Build Consensus — Move from having less agreement among stakeholders to having more.
Initial concept input
A group-intelligence elicitation session starts with an input tool that gathers concepts from a local or distributed team of knowledge workers, AI agents and domain experts. In this example, team inputs are "bucketed" in SWOT categories, but the categories are completely flexible and can be different for every exercise.
Multi-scale rating tool
After concepts are received from team members, subsequent exercises organize, rank, rate and prioritize the items. Each discussion item can be ranked by multiple criteria. The contributions of humans and AI agents are interleaved so as to strengthen and validate outputs.
Check-off exercise
The multi-check tool lets users prioritize items they have defined in a previous exercises.
The Science of Human/AI Collaboration
Under the covers, AI-enhanced problem solving exercises are assisted with patterns drawn from cognitive science and Group Intelligence research. One model for shared cognition is based on Thinklets, small units of useful, repeatable collaborative thought process. Each block in this diagram represents a Thinklet that humans and AI can share.
(A Conceptual Foundation of the ThinkLet Concept for Collaboration Engineering; Briggs, et. al.)
This is a partial list of thinklet collaboration patterns. A group intelligence session comprises a set of thinklets organized into a generative flow with a tangible outcome. Thinklets are: "Named, packaged facilitation strategies that create predictable, repeatable patterns of collaboration."
(Collaboration Engineering: Foundations and Opportunities, Gert-Jan de Vreede, et. al.)
AI-augmented Group intelligence exercise
Shown is an AI guided cybersecurity attack and penetration exercise. The same process flow could be applied to marketing, product design, risk management, decision support, health care, etc.
Different thinklet shared thinking patterns can be used at each step in the collaboration process.
Group Intellignce software simulates many of the qualities of real-world collaboration spaces but without the "group grid-lock" and heard mentality. Shown above: Pixar collaboration enclave.
Body of work
The approach to "virtual mindmeld meetings" described above is based on my work with senior facilitators and collaboration experts over the past 10+ years. With the help of opensource software and AI dev tools, advanced collaboration software solutions can be built for a wide range of team problem-solving exercises at affordable costs.
Companion sites
Here's a companion info page, which looks at knowledge-exchange platforms that scale to very large communities:
Group Intelligence software focuses on "augmented" workgroups. In contrast, Topic Community plaforms are for larger problem-solving enclaves, i.e., different solutions but with lots of overlap.
Thanks to Paul Adams, Bob Briggs, Derek Leebaert, Karen Chenette, David Tobey, David Boje and the many other researchers, collaboration experts and shared thinking advocates that I have communed with over the years.
Apple's campus in the Bay area is a collaboration spaceship. If your org can't afford this sort of workgroup tool, Group Intelligence software is the next best thing :)
Please see companion sites on analytics-driven collaboration enclaves and more:
My Human+AI Knowledge Communities
Let's brainstorm exponential opportunities and applications:
Back to top