Model Census

JavaScript isn't enabled in your browser, so this file can't be opened. Enable and reload.

Model Census

In order to increase transparency of developing/deployed models, we are conducting a brief survey.

Please provide a separate submission for each of the models you currently support. If you support a model with other team members, please coordinate with them on the submission so that you don't have to duplicate effort.

Email *

Record my email address with my response

Model Name *

Maintainer(s) *

Team *

Repository *

Development Stage *

In early ideation or design

In POC or experimental development (not deployed yet)

Preparing for production (not deployed yet)

Already in-operations/deployed in the market or business unit(s)

Previously in-operations, but paused for refinement

What is the model's objective? (i.e., "Provide recommendations") *

What are the business Key Performance Indicators (KPIs) that this model supports? If unsure, leave blank.

Does your model use any of the following (Select all that apply) *

Static Algorithms: Rules or knowledge-based logic systems

Statistical Analysis (e.g., A/B tests, interval estimation, etc.)

Shallow ML algorithms (e.g. Linear regression models, tree based models, SVM, nearest neighbors, etc.)

Deep ML algorithms (e.g., Multi-layer Neural Networks)

Knowledge Graphs (e.g. Spatial analysis, Social networks, Path optimization, etc.)

Image processing or object recognition (e.g. Computer vision, image recognition, image segmentation, etc.).

Natural Language Processing (NLP)

Recommendation or Content generation (e.g. Recommendation engines, Chatbots, etc.)

Adversarial techniques (e.g. GANs, etc.)

Reinforcement learning (e.g. Q-learning, MDP)

Risk or Anomaly detection

Other:

Required

What is the autonomy level of the model? *

The system replaces a decision that would otherwise be made by a human; is autonomous/self-learning (human-out-of-the-loop)

The system decisions are overseen and can be overridden by decision-maker when needed (human-over-the-loop)

The system is used only to assist a decision-maker, but human has the last word (human-in-the-loop)

Provide insight (informational only)

Other:

Retraining Frequency *

Self/continual learning

Daily

Weekly (<1 week)

Monthly (>1 week but <1 month)

Semi-annually (>1 month but <6 months)

Annually (>6 months but <12 months)

Rare (>12 months)

Automated based on drift-detection

Not sure

Other:

How is this model validated for production after training? *

What metrics for the model are being tracked? How and where are the metrics tracked? *

What is the model serving/training pattern? *

Batch Serving / Offline Training

On-Demand Serving / Offline Training

Batch Serving / Online Training

On-Demand Serving / Online Training

Other:

What technology is the model run on when serving predictions/recommendations/etc? *

Apache Airflow

AWS Batch

AWS Lambda

Kubernetes

ECS

Cronjob on EC2

Other:

What type of monitoring (If any) is used to observe this model's behavior in production? *

Do internal users/stakeholders interact with this model? *

Yes

Not sure

Do external users/stakeholders interact with this model? *

Yes

Not sure

If yes to the above, please describe the interaction briefly.

Does the model evaluation or monitoring process include bias detection for user groups.
*

Yes

Other:

Please share any additional details or links to documentation that you think could provide helpful context on this model:

Submit

Clear form

Never submit passwords through Google Forms.

This content is neither created nor endorsed by Google. - Terms of Service - Privacy Policy

Does this form look suspicious? Report

Forms