AI FAQs
We try to consider potential biases with the AIs generated content. We need to understand the samples used in training these models. If the model has been trained in America, there's a chance the outputs may reflect biases due to different social norms compared to the UK. How do we identify and address these?
We use two methods to monitor for bias. Firstly, we consider correlational data that looks at the consistency of pupil scores over time in comparison to scores derived entirely from human judging. Secondly, we ask teachers to make a small number of decisions so that we can monitor how the decisions made by teachers reflect the decisions made by the AI. Analysis of these data ensures we can monitor bias.
How will you be monitoring the Ai to ensure the outputs are accurate and correct? What communication is there for staff to ensure they consider this in the results?
We have created an AI disagreements report for staff to ensure they consider the accuracy of the AI decisions.
Can I turn off AI judging for specific students?
It is not possible to turn AI judging off for specific students. If you have specific students whose work cannot be AI judged, then you have two options. One: remove their work from the assessment. Two: use 100% human judges.
You can see our recommended letter to parents here. We recommend that you inform parents about the use of AI in these assessments, but you do not need positive consent from parents.
How do I ensure the data that is sent to America is secure?
We shall only transfer personal data outside of the UK and EU if we have implemented Appropriate Safeguards to ensure the protection of the Personal Data in the destination territory, in accordance with the applicable Data Protection Legislation. The data may be transferred to the Anthropic servers in the United States but Anthropic ensures it benefits from an adequate level of data protection by relying on Adequacy decisions and Standard contractual clauses. OpenAI encrypts all data at rest (AES-256) and in transit between our customers and us and between us and our service providers (TLS 1.2+), and uses strict access controls to limit who can access data. Both Anthropic and OpenAI are signed up to the data bridge.
We use two methods to monitor for bias. Firstly, we consider correlational data that looks at the consistency of pupil scores over time in comparison to scores derived entirely from human judging. Secondly, we ask teachers to make a small number of decisions so that we can monitor how the decisions made by teachers reflect the decisions made by the AI. Analysis of these data ensures we can monitor bias.
How will you be monitoring the Ai to ensure the outputs are accurate and correct? What communication is there for staff to ensure they consider this in the results?
We have created an AI disagreements report for staff to ensure they consider the accuracy of the AI decisions.
Can I turn off AI judging for specific students?
It is not possible to turn AI judging off for specific students. If you have specific students whose work cannot be AI judged, then you have two options. One: remove their work from the assessment. Two: use 100% human judges.
You can see our recommended letter to parents here. We recommend that you inform parents about the use of AI in these assessments, but you do not need positive consent from parents.
How do I ensure the data that is sent to America is secure?
We shall only transfer personal data outside of the UK and EU if we have implemented Appropriate Safeguards to ensure the protection of the Personal Data in the destination territory, in accordance with the applicable Data Protection Legislation. The data may be transferred to the Anthropic servers in the United States but Anthropic ensures it benefits from an adequate level of data protection by relying on Adequacy decisions and Standard contractual clauses. OpenAI encrypts all data at rest (AES-256) and in transit between our customers and us and between us and our service providers (TLS 1.2+), and uses strict access controls to limit who can access data. Both Anthropic and OpenAI are signed up to the data bridge.
Updated on: 29/04/2025
Thank you!