> For the complete documentation index, see [llms.txt](https://docs.nested-knowledge.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.nested-knowledge.com/review-workflow/screening/exclusion-based-screening/robot-screener.md). # Robot Screener Learn how to use Robot Screener to fill one reviewer seat in Dual Screening. **Robot Screener** fills one reviewer seat in Dual Screening. It learns from your team’s adjudicated Exclusion-Based Screening decisions. Unlike **Smart Screener**, it doesn’t answer defined eligibility questions. It learns patterns from your team’s Advance and Exclude decisions. It then assigns an inclusion probability to each new record. Use it when an automated reviewer can fill one seat. In Dual Two Pass Screening, it works only during Abstract Screening. Full Text Screening remains human. In Single Screening, use the [Screening Model](/review-workflow/screening/exclusion-based-screening/screening-model.md) to prioritize your queue without an automated reviewer. ### Set Up Robot Screener {% hint style="info" %} Confirm your nest uses **Exclusion-Based Screening**. **Robot Screener** does not work with **Criteria-Based Screening**. {% endhint %} **Robot Screener** needs 50 adjudicated screening decisions, including 10 advancements or inclusions. After you turn it on, it trains on further adjudicated decisions. {% stepper %} {% step %} #### Confirm Dual Screening On **Nest Settings**, confirm your nest uses Dual Screening. Robot Screener isn't available in Single Screening. {% endstep %} {% step %} #### Meet The Threshold Two human reviewers must screen and a third human reviewer must adjudicate 50 decisions. At least 10 records must be advanced or included. You can’t turn on Robot Screener before meeting this threshold.
Robot Screener setting unavailable before the screening threshold is met.

Robot Screener remains unavailable until your nest meets the screening threshold.

On **Nest Settings**, select **Robot Screener** to view your progress toward the threshold.
Robot Screener threshold progress in Nest Settings.

Nest Settings shows progress toward the Robot Screener threshold.

See [**Screen Records**](/review-workflow/screening/screen-records.md) for manual screening guidance. {% endstep %} {% step %} #### Turn On Robot Screener On **Nest Settings**, turn on **Robot Screener**. When the threshold is met, select **Enable Robot Screener**.
Enable Robot Screener confirmation dialog.

Confirm that you want to enable Robot Screener.

Robot Screener begins processing records in the background. {% endstep %} {% step %} #### Continue Screening One human reviewer continues screening. Robot Screener provides the second decision. Each record moves to adjudication after both decisions are available. Human adjudicators can view disagreements between the reviewer and Robot Screener.
Human and Robot Screener decisions shown for adjudication.

An adjudicator can compare human and Robot Screener decisions.

{% endstep %} {% step %} #### Filter And Download Robot Screener Decisions To filter Robot Screener decisions, open **Study Inspector**. Select **Add Filter**, then select **Reviewer Screening (Standard)** or **Reviewer AB Screening (Two Pass)**. Select a status, then select **Robot Screener** as the user. To scan decisions, add the **Reviewer Screened By** column. It lists every reviewer and their screening decision. To view a decision, select a study, then turn on **Screen** or **AB**, and **Adjudicate**. Select **Download** to export the filtered studies.
Robot Screener decisions filtered in Study Inspector.

Study Inspector filters records by Robot Screener decisions.

{% endstep %} {% endstepper %} ### How Decisions Appear Robot Screener doesn’t assign exclusion reasons. Its decisions display as **Advance**, **Include**, or **Robot Excluded**. It never finalizes a record on its own. A human reviewer provides the second decision. A human adjudicator resolves any disagreement. Reviewers can’t view Robot Screener decisions during screening. Human adjudicators can view them during adjudication. ### Implications For Your Workflow Turning on Robot Screener changes more than who screens. Consider these workflow changes before you rely on it. After it’s active, you need one human reviewer. Robot Screener fills the second reviewer seat. Robot Screener doesn’t assign a reason. Records it excludes show as **Robot Excluded**, not a specific exclusion reason. Account for this when reporting a PRISMA-style tally by reason. See [**Exclusion-Based Screening**](/review-workflow/screening/exclusion-based-screening.md). A human reviewer provides every record’s second decision. A human adjudicator resolves disagreements. Robot Screener never finalizes a record alone. See [**Adjudicate Records**](/review-workflow/screening/adjudicate-records.md). Because Robot Screener trains on each adjudicated decision, early inconsistencies can affect later decisions. Consistent screening matters most when the model has limited training data. In Dual Two Pass Screening, **Robot Screener** replaces a reviewer only during Abstract Screening. Full Text Screening remains human. ### Review Robot Screener Performance Review predictions and cross-validation metrics each time the model trains. On **Nest Settings**, select **Screening → View Screening Model**.
Screening Model predictions and cross-validation metrics for Robot Screener.

The Screening Model shows predictions and cross-validation metrics for Robot Screener.

Predictions The **Predictions** tab charts Included or Advanced, Excluded, and Unscreened records. The x-axis shows inclusion or advancement probability from 0 to 1. Performance varies by nest because the model trains on your team’s decisions. The optimal threshold is the probability at or below which the model recommends exclusion.
Cross Validations The **Cross Validation** tab shows evaluation metrics, including recall and accuracy. Use these metrics with your protocol and human review process when deciding whether to use **Robot Screener**.
Understand Model Limits The model learns from your team’s adjudicated decisions. Limited include or advance examples, incomplete metadata, or information available only in full text can reduce its reliability. Use **Robot Screener** to fill one reviewer seat, not to make a final decision. A human reviewer and a human adjudicator remain part of every workflow. Model performance varies by nest. Review your model’s metrics before relying on its decisions. See the [Screening Model model card](/ai-transparency/model-cards/screening.md) for evaluation details, input data, and limitations.
### FAQs
What happens before the 50-decision threshold is met? **Dual Screening** uses two human reviewers and a human adjudicator until you turn on **Robot Screener**. It can’t fill a reviewer seat before the threshold is met.
How do I improve Robot Screener performance? Adjudicate consistently and review model metrics after training. More adjudicated decisions provide more review-specific data. They don’t guarantee a particular performance outcome.
Can Robot Screener be turned off after it's turned on? Yes. Turn it off any time on **Nest Settings**. When you turn it back on, it trains a new model from current screening decisions.
Can I use Robot Screener with Criteria-Based Screening? No. It’s available only with **Exclusion Reasons**. **Criteria-Based Screening** uses **Smart Screener** instead.
### Next Steps
:clipboard-check:
Screen Records		Screen alongside Robot Screener in Dual Screening.Screen Records
:scale-balanced:
Adjudicate Records		Resolve disagreements with a human adjudicator.Adjudicate Records
--- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation by asking a question. Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters: ``` GET https://docs.nested-knowledge.com/review-workflow/screening/exclusion-based-screening/robot-screener.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.