Using AI tools in research

Artificial Intelligence (AI) is revolutionising and disrupting various fields - including the research space - where AI tools are increasingly being utilised to enhance research capabilities and streamline processes.

This information is designed to start you on your journey to learn more about this exciting technology and give you some confidence when exploring these tools.

What is AI?

Appraising AI tools

Using AI in research

Using AI for study

University statement on AI in research

What is artificial intelligence?

Artificial intelligence (AI) refers to computer systems that are able to perform tasks that usually require human intelligence. There are different types of AI in use today.

Artificial narrow intelligence and Artificial general intelligence

Artificial narrow intelligence (ANI)

ANI refers to systems that are programmed to perform a single task, using machine learning techniques such as supervised learning and reinforcement learning.
Examples include image recognition in self-driving cars, recommendation systems when shopping online, or text prediction apps.

Artificial general intelligence (AGI)

AGI is a theoretical type of AI that is able to perform any intellectual task in any situation, much like a human.

Machine learning

Machine learning refers to a range of techniques to build systems that have the ability to learn and improve from experience, without being explicitly programmed to do so.

Two common examples of machine learning are supervised learning and reinforcement learning.

Supervised learning

This is where a program is given data that has been labelled. From this, the program starts to recognise patterns, allowing it to predict or classify new information. Examples include image recognition, where programs are trained to recognise cats.

Reinforcement learning

This involves giving the program feedback each time it performs a task, so that it learns from doing things correctly.

Generative artificial intelligence

Generative AI is a type of artificial narrow intelligence (ANI) where the program creates new content – either text or images – based on the data that is has been trained on.

A recent example of generative AI is ChatGPT. ChatGPT has been trained on a massive amount of text from digital resources to recognise patterns in words and sentences. It provides responses to prompts by predicting the next likely word in a given context.

You may also have heard of DALL-E, which was trained on a large set of image and text pairs to create new images based on textual descriptions.

Appraising AI tools

When using any new digital tool or source of information, it is good practice to think critically about why you are using it, how it can help you, and what its limitations are.

Purpose of the tool

Look for information provided by the developers about the tool you want to use.

What do they say about why the tool was developed?
How do they envisage it will be used, or how it will help people?
Does this match what you want to do?

Look for information about how others have used the tool.

What comments do they have about what it is good for?
What its strengths and weaknesses are?
Explore the tool yourself. Enter different prompts and see what happens. Test it against something you know really well or are starting to learn.

Accuracy and trustworthiness

A useful expression in computer science is “garbage in, garbage out”. Tools are only as good as the data used. If programs are trained on incomplete, inaccurate or biased data, then the output will also be incomplete, inaccurate or biased. The output is also influenced by the quality, accuracy, and thoroughness of the training that the program receives.

Look for information provided by the developers about the tool you want to use.

Does it include a description of the data set and the training methods used?
Are there any obvious gaps or biases?
What do other tech or subject experts say about the tool and its outputs?
Would this impact the reliability

Relevancy

It can take a long time to train AI programs and a long time for new datasets to be incorporated into its training.

Does the program indicate when it was last updated and how current its dataset is? Does this impact the relevance of the tool for your requirements?

Legal considerations

As with any digital tool, it is good practice to check any terms of use.

Look for the following:

Restrictions on use

Are there limitations on who can use the tool? E.g. persons under 18 years of age.
Are the limitations on how the outputs can be used? E.g. can it be used for educational purposes, but not commercial?

Intellectual property

Who owns the outputs produced?
What can you do with the outputs? Can you share it, re-use it, edit it?
Is there a requirement to provide an acknowledgement when using the output?

Privacy and security

Do you need to provide any personal information to use the tool?
Where is this information stored? Who has access to it? What can they do with it?

Social and ethical considerations

As well as the relevance, trustworthiness and use-limits of AI tools, you may be interested to think about some of the social and ethical aspects of AI tools.

For example:

Use of copyrighted material when training AI programs.
Equity of access (e.g. where users need to pay to use the tool, or where a minimum level of internet access or computing power is needed to use the tool).
Environmental impact of developing and running large AI programs.

Using AI tools in research

Whilst AI tools can offer several benefits for researchers, there are significant limitations and risks to these tools that need to be carefully considered. AI tools are rapidly evolving - as is the conversation with publishers and governing bodies around the ethical and responsible use of AI in research.

Ethics and privacy

Ethical concerns such as privacy, lack of regulation and data security are central to the conversation around the use of AI in research. These concerns need to be recognised as you explore AI tools to ensure that your research is conducted in a responsible way, with transparency around use of AI tools.

Learn more: Australia’s Artificial Intelligence Ethics Framework

Bias and data sources

AI tools are only as good as the data they are trained on – if that data contains biases, then the tool can also be biased. This is one of the primary concerns in using AI tools in research as it can lead to inaccurate or misleading findings that could negatively impact research outcomes. Interrogate the data source of any AI tool you plan to use so you can be sure of where the information is coming from and what the implications are for the integrity of your research.

It is important that you consider copyright and the University’s Intellectual Property Policy when using AI tools to both contribute content and when using AI generated material for university purposes. Content created by AI is not protected by copyright and may incidentally infringe the rights of others.

Need help? Submit a Copyright for research query via ServiceOne.

Exploring AI tools

There are numerous AI tools already available that researchers might like to explore.

These tools can be used for a range of different tasks such as finding academic papers, summarising academic papers, connecting ideas, and evaluating the credibility of academic papers (Elicit, Scholarcy, Scite, Semantic Scholar). Other tools focus on mapping literature on specific topic/s (Connected Papers, Inciteful, Litmaps, ResearchRabbit), or can help a researcher identify relevant journals to consider in their publishing journey (B!SON).

Many of the AI tools used for research purposes are free, whilst some charge a subscription or one-off fee to access additional features. While paid tools often provide advantages in terms of functionality and support, high quality free tools are also available – be sure to evaluate your specific requirements, budget and weigh up the available options to determine if a tool is going to suit your needs.

The tools we have mentioned on this page are only a few examples of the tools that are available – there are more complex tools, ones that are simpler, and new tools rapidly being developed. Any AI tool used in research requires rigorous appraisal and needs to be evaluated within the context of appropriate conduct at Flinders University and in the publishing landscape.

AI and research grants

Researchers need to exercise caution when using AI tools to develop grant applications, or when participating as a peer reviewer. Both the Australian Research Council (ARC) and the National Health and Medical Research Council (NHMRC) have released policies providing guidance for researchers in relation to the use of AI tools, with the potential for other funders to follow suit.

ARC: Policy on Use of Generative Artificial Intelligence in the ARC’s grants programs

NHMRC: Policy on Use of Generative Artificial Intelligence in Grant Applications and Peer Review