NEW WAYS OF ORGANIZING · SHORTS  01

Welcome to our new initiative: concise insights into how we can improve our collective action through new forms of collaboration, governance, ownership, and value distribution. While decentralized autonomous organizations (DAOs) can be part of this, we adopt a broader perspective on collective action with this initiative.

Getting quality results from AI agents

The use of AI and AI agents is on the rise; both in society at large and within organizations. For organizations this means a new type of workers and transaction parties: non-human entities.

However useful AI agents may be, we humans face a major challenge: how do we maintain control over them?

Because AI agents behave non-deterministically, they can—and will—produce unexpected results. How do we increase the likelihood that AI agents meet our needs, knowing that they rely on opaque algorithms?

Given the way AI agents generate results (quickly, comprehensively, and in a manner that is flattering towards us), we risk falling into the trap of complacency (I may be speaking for myself here): we issue a brief, vague instruction (prompt) and accept the outcome without verifying or testing it.

To make AI agents truly work for us, we must approach their mission as a ‘principal-agent’ issue, with us humans fulfilling the role of ‘principal’ (the party issuing the mandate). This role could even be performed by a steering group comprising multiple experts.

In such a setup, the AI ​​agent’s mission receives the necessary attention and input from various disciplines. Additional measures might include a mission input report, a debriefing report, and thorough testing and evaluation of both the result and the mission itself.

A mission input report might include the prompt, the assigned tools, and the configuration settings (such as tone of voice, temperature, and model). A debriefing report—featuring a timeline of the tools used and sources consulted—provides insight into how the AI ​​agent acted during the mission.

Decentralized Autonomous Organizations (DAOs) - Mechanisms/tools

Simply pressing an 'Approve' button on an AI agent's output—as a way to keep 'humans in the loop'—will all too often lead to disappointment. To obtain high-quality results from AI agents, one must meet the criteria of a ‘good’ principal. Ideally, a group of experts acts as the principal. DAO-mechanisms/tools—such as providing shared, reliable information between decision-makers—can help put this ‘good’ principal role into practice. See for more info https://v0-how-to-dao-it-website-with-wiki.vercel.app

Want to exchange ideas or need information or help?

Contact us at +31 6 53 42 42 93 or info@newwaysoforganizing.com. Follow New Ways of Organizing on LinkedIn: linkedin.com/company/new-ways-of-organizing

#DAO #Collaboration #Governance #Organizing #Trust #Tools
#AI-agents #Control

Street art as an example of a (digital) common

Human artists create street art in public spaces—free for anyone who encounters it.

By Lemak, Bristol, UK