Model Cards for AI Model Transparency

At Salesforce, we take seriously our mission to create and deliver AI technology that is responsible, accountable, transparent, empowering, and inclusive. These principles ensure that our AI is safe, ethical, and engenders trust. We commit ourselves to:

• Asking not only can we do something but also should we do it, before building any AI features.
• Providing clear and understandable explanations of our AI’s predictions or recommendations. When an AI function makes a prediction or recommendation, we will provide explanations of the model.
• Outlining the intended applications and potential impacts of AI capabilities.

One key element of our commitment to Transparency is the production of model cards. Based on pioneering work done by Google, model cards seek to standardize documentation procedures to communicate the performance characteristics of trained machine learning (ML) and artificial intelligence (AI) models. Think of them as a sort of nutrition label, designed to provide critical information about how our models work -- including inputs, outputs, the conditions under which models work best, and ethical considerations in their use.

Providing this level of transparency is intended to help developers, customers, journalists and industry analysts, policy makers, advocacy groups, and consumers better understand the impact of our AI on individuals, communities and society.

The following model cards are currently available for reference:

Salesforce Research
CTRL
Detector for CTRL
AI Economist: Foundation (Simulation Framework)

Marketing Cloud Einstein
Send Time Optimization
Messaging Insights

Sales Cloud Einstein
Call Coaching Language Model

Currently in development are cards for:

• Sales Cloud Einstein Lead Scoring & Opportunity Scoring
• Marketing Cloud Einstein Engagement Frequency and Engagement Statistics
• Optical Character Recognition
• ...and many others!

We look forward to sharing these with you and welcome your feedback, reactions and questions.