You can only use GenAI if permission is specifically given in an assignment brief. If you are unsure you should speak to the module convener. We encourage people to be transparent about their use of GenAI tools so that they can use them effectively without risking the academic integrity of their work. Remember, unauthorised use of GenAI in assignments could breach the university's regulations on academic misconduct.

This is the process of structuring a prompt to ensure it is properly understood by a GenAI tool. There are a few techniques you can use to write more effective prompts and get better outputs. These tips might not work on all GenAI tools.

• Write very clear, precise instructions using natural language (how you would speak to another person).

• Use subject specific terminology, where appropriate, to improve the relevance of the tool's response.

• Specify the length and depth you want a GenAI tool to go into with its response.

• Ask the GenAI tool to respond in a particular style. For example, you could ask it to respond in an academic style or informally. Alternatively, you could ask it to generate content for an email or social media post.

• Specify the language in which you want the GenAI tool to respond. Remember there are differences between UK and US English.

• For difficult concepts it can help if you break down your question into a series of prompts. This might be as a logic sequence, or as an ongoing conversation to build a more thought through response.

• Ask the GenAI tool to generate its own solution to a problem first, then compare it to your solution. The tool is then more likely to identify any issues and suggest corrections to the solution.

There are a number of GenAI tools, such as DALL-E, Adobe Firefly, and Midjourney, which can create images. Some require payment to access more advanced options or generate multiple images. The outputs from these tools will be reflective of the dataset that it draws from. This means that images can reflect biases and stereotypes. Some tools can also be prone to misunderstanding some prompts, misspelling words, or having noticeable errors in the image. For example, the image below was created with the prompt "can you create an image of a 5 librarians standing in front of a book shelf please?" As you can see, there are 7 people pictured in the image.

There are a number of tips you can use to try and improve images made by GenAI.

• If you are using AI generated images you should always state this explicitly and include it in the ALT text description of the image.
• When writing a prompt for an AI generated image start with a brief description of the basic premise of the image you want to create. For example, you might write "an image of a sunrise over a city".
• You can then define the artistic style of the image. For example, you might say "in the style of a water colour painting", or "as a photo realistic image".
• Some tools allow you to specify the image ratio and size. For example, you can ask it to create a square image, a wide image, or a tall image. A wide image will be in landscape and a tall image will be in portrait.
• You can also add specific details to your prompt. For example, you could add "the city should have modern buildings with a mountain range in the background".
• As a general rule, try to keep things simple. The more details you add to a prompt the more likely you are to get unexpected results.
• Many GenAI tools will allow you to refine an image once it has been created. However, it will most likely create a whole new image adding in the additional prompts rather than refining the original image. 
• If you do not get a satisfactory output then tweak and change the wording of the prompt. It is worth practicing to see which prompts get the best results.

Using the examples provided above, here is an AI generated image from the prompt "can you create an image of a city at sunrise in the style of a water colour painting, in a square format, please. The city should have modern buildings and have a mountain range in the background".

GenAI tools can be prone to hallucinations. This is when they generate information that is factually incorrect or incomplete. It is also difficult to verify what source a specific GenAI tool is drawing from. Sometimes asking for references on a topic can result in the GenAI tool making up resources. There are a few ways you can verify the reliability of outputs from generative AI tools.

• Don’t just rely on the outputs from a GenAI tool. You should also use other resources to build your knowledge.

• Can you corroborate the information? Check other authoritative resources to see if they match the output of the GenAI tool. Be sure to check if there is an alternative perspective or argument that might not have been covered by the generative AI tool.

• Check the data set that the GenAI tool has been trained on. Does it use good quality resources to generate its outputs? If it is unclear what data it is using then you should carefully consider its reliability.

• If the GenAI tool quotes or references any resources make sure you check their authenticity and the GenAI tool has interpreted them correctly.

• Does it contain any biases in the outputs or dataset? The outputs will only be as good as the data the GenAI has been trained on. If certain perspectives or types of resources are missing, then the outputs may reflect this and contain biases.

Before using an output from a GenAI in an assignment you should consider its academic quality, whether it is evidenced, and if it would be appropriate to use in an academic assignment. If you are unsure, then you should contact your module convener. 

The website Cite Them Right contains general guidance on the use of Generative AI in academic work. For guidance on how to reference GenAI in specific referencing styles visit How to reference a generative AI output on Cite Them Right.

Privacy – Be careful what personal information you put into a GenAI tool, especially if you are unsure how the GenAI tool might use that information. This does not just include information about yourself, but information about other people where you have not has permission to input their data into an GenAI tool. For academic staff it is especially important not to upload student assignments into GenAI tools. Some GenAI tools do have a privacy policy, or outline how they may use inputs to help train the AI. Others might share the data they collect with third parties. Ultimately, you need to ask yourself if you would share the information with a stranger where you have no idea how they might use that information.

Copyright – You should avoid uploading documents and information that is copyright protected. For example, journal articles which aren’t Open Access shouldn’t be uploaded into GenAI tools. Even if you are the author, often the publishers will own the copyright. This also includes funded research or proprietary information. Even if you have created it, you might not own the copyright. You should also be mindful of upload course materials and learning resources into GenAI tools. Typically the resources creator or the university will have ownership of these materials and, unless permission is specifically granted, it might not be appropriate to upload them.

Digital equality and inclusivity – AI tools have the potential to support learners with disabilities or when English isn’t their first language. It can help with proof reading, translation, organising notes, and creating revision timetables. However, we need to remember it is only accessible for those with access to reliable internet, with a computer or smart device, and in some instance, the money to afford a subscription. We need to mindful when using GenAI in education that all learners have fair and equitable access to any GenAI tools they are permitted to use. They should also be taught how to use the tools effectively and appropriately.

Environmental impact – GenAI Tools have the potential to come up with solutions on how to better manage the environment and reduce the impact of human activity on climate change. But in turn running servers and systems capable of hosting AI tools can be resource intensive. It is estimated that up to 9 litres of water to evaporate per kWh of energy used at AI data centres. They often have poor water recycling processes so can dramatically affect water supply chains in areas where they operate. There is also the energy required to power the servers. This doesn’t always come from green energy so can have a large carbon footprint.

Employment Practices - Several news outlets (TIME, 2023; The Washington Post, 2023; The Guardian, 2024) have reported that some GenAI companies have been using questionable employment practices. This has varied from paying low wages to staff in developing countries, having long working hours, or exposing them to harmful content. They have also highlighted that often low paid workers are required to do repetitive tasks for long hours, such as data labeling (adding descriptions or labels to information to help a GenAI tool better understand it). Consequently many of these workers have been exposed to harmful content with allegedly little to no mental health support.