Atomi automatically marks responses to short answer questions using artificial intelligence (AI), allowing students to receive instant feedback when they need it most.
When you submit a response, Atomi's AI will grade your response as either; Correct (2 marks), Incorrect (0 marks) or Partially correct (1 mark). You will also see a sample answer or what the answer should include.
What makes an answer correct, partly correct or incorrect?
Atomi's AI auto-marking evaluates your answer in light of its relevance, completeness and clarity.
Relevance measures the extent to which your answer supplies factually correct information that addresses the question. In other words, did you provide the information the question was looking for?
Completeness is a measure of the extent to which you fulfilled all of the instructions or requirements of the question. For example, if the question asked you to "give an example" and you did not include an example, your answer might be relevant but incomplete.
Clarity measures the quality of your writing and how easily the model can understand your response. Your choice of words, the specificity of your language and your spelling, syntax and grammar all impact clarity. We've designed our AI model to be quite forgiving in this regard, but a minimum threshold of clarity is required for an answer to get full marks.
Marks are awarded as follows:
Label | Definition | Marks |
| When an answer has all of the following qualities:
| 2 |
| When an answer either:
| 1 |
| When an answer either:
Generally speaking, incorrect answers do not adequately satisfy any part of the question. | 0 |
How do I access AI-marked short answer questions?
AI-powered auto-marking is being progressively rolled out across Atomi. At this point, Atomi's AI marks most short answer questions across English, Maths, Science and Physical Education in interactive lessons across Years 7-10. AI auto-marking is not available in Atomi Revision Sessions.
As Atomi's AI model is still learning, there are times when it might not be able to confidently decide which mark to give your answer. When the AI model doesn't have a high confidence level to auto-mark, we'll fall back to asking you to self-mark your response.
We're constantly improving the Atomi AI model so these and other subject and revision sessions will be unlocked in the future.
How accurate is Atomi’s AI at marking?
Atomi’s AI is trained and tested on thousands of Atomi questions and actual student responses, with tight supervision by our in-house machine learning experts, academic content creators and high school teachers.
Testing has shown that Atomi’s AI model grades student responses with perfect accuracy over 80% of the time and is within one mark of accuracy 99.6% of the time. This accuracy is comparable to state of the art models in the field of Automatic Short Answer Grading.
In comparison, most students are far more likely to mark themselves incorrectly when self-marking.
If you believe your answer was marked incorrectly, you can log a dispute. The Atomi team uses this information to further improve the accuracy of Atomi's AI model.
As a teacher, can I see what mark Atomi’s AI gave?
Student marks for short answer questions appear in the Marks tab in the lesson progress report, just like any other question type. There is, however, no way to distinguish between a question self-marked by a student or auto-marked by Atomi's AI at this time.
How does Atomi protect the privacy of student data?
Atomi is a Student Privacy Pledge signatory, committed to protecting student privacy.
AI auto-marking occurs entirely within the Atomi platform, with no data sent to third parties. Anonymised student responses to questions may be shared with a third party to improve Atomi’s AI model. However, no data about the student is shared.
You can learn more about the types of data we collect, how we use and store data and your rights in our privacy policy or the Security, privacy and trust section of this Help Centre..