Elemental uses AI responsibly to provide Summational's AI marking feature. Elemental does not use Large Language Models (LLMs) to process Student Data and Student Data is never used by Elemental or our sub-processors for model training.

What is Summational AI Marking?

Summational is designed to provide assessments under controlled conditions, always based in sound.

Teachers can choose to collect each student's answers on paper worksheets, which can then be scanned and uploaded to Elemental to be marked automatically.

These scans are processed by a pre-trained model, not a Large Language Model (LLM). This model performs Optical Character Recognition (OCR), converting the scan into text and layout information so that student answers can be located on each page. The scans are processed by a trusted sub-processor within the UK.

The extracted marks are then presented to the teacher for review, with warnings on any potential marking issues. Teachers are able to override the automated marks and are encouraged to do so as appropriate.

You can learn more about Summational on the Summational product page.

What we do not do

• No AI-generated music. None of the music in Elemental is AI-generated, and it never will be. Every musical example is either out-of-copyright repertoire or new music written by composers and songwriters working in a broad range of styles.
• No generative AI marking. Elemental does not use Large Language Models (LLMs) or any generative AI to mark, comment on, or make decisions about students' work.
• No training on Student Data. No Student Data is used to train any AI model (our trusted sub-processors have made the same commitment).
• No automated marking without teacher review. Marking outcomes are presented to teachers, who remain responsible for assessing students' work and are free to override the automated marks (and are encouraged to do so as appropriate).
• No profiling or automated decision-making with legal or similarly significant effects. With the exception of the Summational AI marking feature, all marking is fully deterministic. No marking is ever influenced by Student Data, such as a student's name, and students are never profiled based on this data.