The background material compiles the knowledge base for AI recommendations and supports education providers and teachers in developing AI competence and utilizing AI locally.

The background material covers various themes and should be used as a whole.

Artificial intelligence (AI) processes and generates information, but it does not have human-like understanding or a consciousness. Unlike humans, AI cannot create meaning and interpret information, even though generative AI, in particular, mimics human-like behaviour in its output. AI models are based on machine learning, algorithms and the data they are trained on. AI lacks the human capacity for ethical or moral consideration – it has no conscience, emotions, or empathy. For the aforementioned reasons, responsibility for using and interpreting information produced by AI always rests with the user, a human.

The legal, safe, reliable and fair use of AI in education requires its users to understand what AI is, what kinds of misinterpretations AI can make and what types of bias can occur in AI outputs.  

The ability to recognise AI misinterpretations and bias, as well as to prevent discrimination, is a key part of the AI capabilities of educational institutions. However, it is nearly impossible to reliably assess the trustworthiness, misinterpretations or bias of AI based solely on its outputs. The reliable identification of the misinterpretations and data bias of AI is dependent on the operation of the AI systems being made transparent and understandable. 

Currently, there are no universally functioning methods or standards for the automatic identification of AI bias and misinterpretations, as the use of AI is highly contextual and dependent on the subject matter. While methods and applications have been developed for detecting data bias and for validating AI training data and models, using these methods and applications requires extensive specialised expertise. Particularly when using AI-based learning analytics, educational institutions should require the providers of AI systems to provide reliable analysis of their AI’s algorithmic and data biases so that any bias can be identified, enabling accurate interpretation of AI-generated outputs and outcomes, and ensuring that discrimination is effectively prevented.

Misinterpretation: In the context of AI, a misinterpretation can refer to a) AI algorithms produce incorrect results or outputs (‘faulty inference’), which is most often caused by the AI model and algorithms, or b) the user misusing or misinterpreting AI outputs.

Bias: AI bias refers to AI systems systematically generating distorted results and outputs. This may be due to the data used, the operation of the algorithms or intentional design choice. For example, the data used to train the AI may have been compiled from another context or population and applied to the current situation and population. On the other hand, bias may also arise as a result of user interpretations. Biases in artificial intelligence can lead to discrimination and exacerbate structural inequalities and inequities.

Identifying AI misinterpretations and bias requires users to possess good AI skills, which include not only actual algorithmic thinking but also so-called multiliteracy and critical thinking skills. These skills are an important foundation for utilising AI effectively in knowledge creation, reasoning, analysis, and other AI-based outputs.  

AI skills are often new to us, and learning them can be a shared journey for teachers and learners. In addition to the learning of AI skills, this section also supports educational institutions and users of AI in assessing the reliability and accuracy of AI and, in particular, the fairness and ethics of AI and its use. 


Updated