Skip to Main Content
QUL logo

Artificial Intelligence

Artifical Intelligence and the research process

Evaluating Output

Before using generative AI, consider whether the task you hope to complete can be accomplished successfully by AI, and if so, what AI tool is best suited to the task. ​

  • This includes considering the potential shortfalls of a tool, such as what bias impacts the AI model, whether it was trained on incomplete, inaccurate or unrepresentative datasets, or has a limited or outdated scope of information. ​

The content created by generative AI, such as ChatGPT and DALL-E, need to be evaluated just like any other source; however, there are unique issues involved with generative AI that require its outputs to be carefully evaluated.

For example, generative AI can:​

  • Be influenced by bias, and provide biased information​

  • Be susceptible to "hallucinating," or making up information, content, or sources​

  • Make up convincing looking sources, or attribute information falsely to an existing resource​

  • Use inaccurate and outdated data ​

  • Produce irrelevant results​

  • Predict, rather than think

Generative AI cannot actually think or understand information or the relationships between ideas

Related Resources