Following established logics for apps, AI tools are typically offered in combinations of free, freemium and premium options, each with different levels of access and different cost structures.

Paid access may grant faster outputs or more powerful models, or being able to access the service without rate limits.

Consider which level of access may be required to effectively use the tool for a given educational application and what strategies may be used to try to ensure fair use across a diverse cohort of students.

For example, can you require everyone to use only the free version and its functionalities? Should everyone purchase a particular level of access that is clearly identified as a manageable cost associated with undertaking the unit? Can the use case be structured so that receiving a faster output via a paid version does not necessarily disadvantage someone using a free version to accomplish the same task?

What levels of technical and content knowledge are necessary for all students to possess in order to equitably use a given tool/platform?

What support needs to be provided to bridge knowledge gaps across diverse student cohorts?

Some students may already be literate in the forms, approaches and understanding of certain tools or models - and how might that may advantage or disadvantage them.

There are technical and physical barriers to entry for incorporating some technologies into educational deliveries - including specialist hardware and familiarity with software environments.

There are specific models that can be run locally - either on GPUS or on apple silicon. This can be a way of bypassing paid services, or rate limitation - but does come with technical and hardware barriers. In many ways engagement with AI models on a local machine/server is preferred as this controlled environment mitigates all some concerns regarding data privacy, training etc - but it opens up considerations about technical approach, and the resources required to run these.

Engagement with tools is not without effort - time is required to understand the tool, approach, interface, and conventions. This cognitive load can impact on learning experiences and student wellbeing. If learning is structured with multiple platforms and tools across a unit - students will need to retain working knowledge of these tools. That is, diversity, range, and types of tools employed in a unit should be carefully considered from the perspective of variations in their interface, approaches, systems and conventions.

Having engaged with a tool and developed a working knowledge of its processes - a student should reasonably expect to be able to re-deploy this at a later date - that is, coordinating and planning for consistency of tools across units, year levels, courses, faculties has benefits. Selection of tools should consider student experiences in other units - and there may need to be consultation around a set pathway through and appropriate scaffolding of specific AI tools across areas of study within courses.

AI tools and their underlying models have embedded, visible and hidden biases. Commercial services may have worked to add guide rails but the biases will still be present based on training data, training method, biases in human trainers or the user group, and intent/purpose of the tool. This means that approaches toward cultural safety are paramount - and selection of tools should consider their predisposition to generate biased, inflammatory or discriminatory content.

Open discussions about bias with students can foster critical engagement with information and AI outputs.