HAX Research Laboratory

Social Impact

Agentic systems have profound ripple effects: they influence how we work, what knowledge is accessible, how power is distributed, and how we make decisions at scale. That’s why we treat societal impact as a design responsibility, not a byproduct. We ask not just What works? but:

Who does this serve?
Who might it exclude or harm?
What are the long-term consequences of deploying this system at scale?

HAX Research Laboratory

Social Impact

Who does this serve?
Who might it exclude or harm?
What are the long-term consequences of deploying this system at scale?

A Framework for Responsible Agent Design

To support teams building agentic systems, we developed a practical, five-part framework—adaptable across roles, from UX designers to backend engineers.

Contextual Inquiry

Design Begins with Understanding

Map the full socio-technical system: who are the stakeholders, what are the workflows, where does agency shift?
Identify power dynamics: What decisions is the AI making or influencing? Who has override authority?
Conduct interviews, not just with users, but with those impacted by system outcomes (e.g., moderators, QA testers, policy teams).

Templates

Agent Impact Map

Intentional Scope

What Should This Agent Do

Define clear boundaries: Where should the agent intervene, suggest, defer, or stay silent?
Prioritize augmentation over automation: Ask how the agent can make users more capable, not redundant.

Templates

Agent Impact Map

Inclusive Cognitive Design

Respect Diverse Ways of Thinking & Working

Design for neurodiversity and multilingualism.
Support different expertise levels—novices, experts, non-coders, etc.
Minimize cognitive overload: surface what’s necessary, when it’s needed.

Templates

Cognitive Load Audit

Foresight & Feedback Loops

Built for Change. Expect the Unexpected

Use speculative scenarios to anticipate unintended consequences.
Include continuous user feedback mechanisms (not just surveys—embedded nudges, annotations, corrections).

Templates

Foresight Canvas

Social Impact

Who does this serve?
Who might it exclude or harm?
What are the long-term consequences of deploying this system at scale?

Social Impact

Who does this serve?
Who might it exclude or harm?
What are the long-term consequences of deploying this system at scale?