Generative Processing refers to learning activities that require the learner to actively construct meaning and output, rather than passively receiving input. The depth of this processing is scientifically categorized by the ICAP Framework, which predicts learning outcomes based on the level of cognitive engagement.

According to Michelene Chi (2009), overt student behaviors can be categorized into four modes, which predict the depth of learning in a hierarchical order (from least to most effective):

• Passive: Receiving information without overt action (e.g., listening to a lecture or reading text).
• Active: Manipulating information physically (e.g., highlighting text or clicking "next").
• Constructive: Producing new outputs that go beyond the presented information (e.g., explaining "why," drawing a diagram, or constructing a sentence from scratch).
• Interactive: Co-creating knowledge with a partner or system that responds dynamically.

Chi (2009) posits that deep learning—specifically the ability to transfer knowledge to new contexts—requires at least Constructive engagement. This is because constructive tasks force the learner to integrate new information with prior knowledge schemas, a process that does not occur during passive reception.

We strictly enforce Constructive engagement to ensure deep learning .

• No Passive Clicking: Users cannot simply select a multiple-choice answer (which is merely an "Active" or "Passive" recognition task).
• Generative Output: Users must type sentences, fulfilling the requirement of producing output that goes beyond the prompt.
• Interactive Feedback: Our AI grading provides dynamic responses to the user's input, elevating the session to the Interactive level.

1. Chi, M. T. H. (2009). Active-Constructive-Interactive: A conceptual framework for differentiating learning activities. Topics in Cognitive Science, 1, 73–105.