Impact Metrics & Formulas
We believe that AI has the potential to positively redefine how humans flourish. We are already seeing that through the emergence of social good organizations that use AI to deliver impact.
Impact is a loaded word. It's meant to define the SROI of funds or grants while simultaneously having no standard criteria for measurement. AI complicates this as the benefits of automation are accompanied by less transparency into the process.
Our framework turns outcomes of HCAI projects into structured, transparent estimates—with equity, safety, and uncertainty built in. Publish numbers you can defend.
Core Metric: Effective Human‑Centered Impact
How much human good did the AI produce?
EHI = People helped × average benefit × (benefit duration) +
Adjustments for retention, model quality, displacement, human+AI synergy, harms, and attribution
Causal Attribution
How much better are outcomes because of AI?
α = (Before‑after with AI) - (before‑after without AI)
+ adjustments for selection and trends
Safety and Harm
How much harm did AI do?
H = EHI * severity‑weighted harm index
Balanced SROI
How much social value per dollar, net of harms and hidden AI costs?
bSROI = (Monetized Benefits - Negative Effects)
/ Total Costs
Reliability
Do quality and usage hold up over time?
R = (Retention(time) + Quality(time))
/ Total Time
Metrics: Equity & Inclusion
Who Benefits?
Inclusion Rate (Intersectional)
Weighted Inclusion (Barrier Weights)
Success Equity (Outcome parity & gap change)
Uncertainty & Sensitivity
How fragile are our numbers?
Confidence Intervals (95%)
Tornado Sensitivity
Scenario Bands