The following article:

• Outlines the differences between the two main AI chatbots: ChatGPT and Google Bard.
• Explores the use of AI as a language expert to improve writing using an academic writing style/nominalisation.
• Shows how to use AI to critique written work and suggest improvements, e.g. grammar correction, spelling, etc.
• Provides ways to use AI to summarise complex events/processes/topics.
• Investigates the opportunities of AI in the NEA, including the initial research, as well as fieldwork synthesis and summaries.

New York Times, January 2023: ChatGPT, the buzzy chatbot developed by OpenAI that is capable of writing cogent essays, solving science and math problems and producing working computer code.

You certainly will have noticed the rise of AI (artificial intelligence) and chatbots. In fact, you will most likely have used them in one way or another. In a previous Geography Factsheet, this author used the same technology in order to help come up with some ideas for a topic!

At the time of writing, ChatGPT is the main disruptor. However, Google has recently released an early version of its AI – Bard. There are also other chatbots emerging, including Jenni AI, Jasper AI, and Writesonic. In educational circles, these chatbots are causing some strong and divided feelings. One view is that, when used in the right way, AI can be a friend to learning and an amazing tool for students, and is not something to be feared. An alternative position is that teachers might be tricked, and that students may present work produced directly from the chatbot while pretending it was created by them.

ChatGPT is a powerful chatbot, built on top of OpenAI's GPT technology. GPT is short for “generative pretrained transformer”. The chatbot can handle a wide range of text-based requests, from simple questions and answers to more complex tasks.

Google Bard, on the other hand, uses Google's Language Model for Dialogue Applications (LaMDA), and can offer real-time responses based on current research pulled from the internet. That means it should be more up to date compared to ChatGPT, which is trained on data prior to late 2021. Remember that both of these AIs don’t understand what they are actually producing. Their outputs are essentially a series of statistical probabilities turned into a sequence of works or phrases.

	ChatGPT	Bard
Creators	OpenAI	Google
Release date	November 2022	February 2023
Data sources	ChatGPT was trained on a massive dataset of text, including Wikipedia, books, articles, documents, and content scraped from the open internet.	Bard was trained on Infiniset, a data set including Wikipedia, documents, and conversations and dialogues from the web. Bard will supposedly search the web in real-time to find the most recent answers to questions and latest research.
Strengths	Speedy response Very easy to use Very good at text summaries and synthesis	Free to use (selected users at time of writing) Up to date, and draws from the web when answering More focused on academia and research
Weaknesses	Makes up references Information can be out of date Premium access requires a monthly fee	Can provide inaccurate information Responses are not as naturalistic

Overall, Bard is designed to improve research and understanding across education, business, and other fields, while ChatGPT is more focused on text functions, including summarising and synthesising. In practice, they are actually quite similar. It’s going to be a good idea for you to evaluate both of them, if possible. Which of the two AIs is your preferred option?

Arguably, one of AI’s super skills is to use the tool as a language expert. It can help you to improve your writing, which can be useful in a range of situations, especially for helping to improve writing long answers.

Let’s start by assuming that you have already written and typed up an answer to an extended writing essay which is level marked.

You can request (prompt) a range of instructions that will potentially help improve the writing quality of your original work. These can include:

• Paraphrase the following: [and then paste a paragraph of your work. It will rephrase it for you]
• Correct/Proofread the following: [and then paste a paragraph of your work, it will correct it for spelling and grammar mistakes – you can also ask it to show the mistakes in that paragraph]
• Bullet point: [then add the topic/research of your choice]
• Give a timeline for: [then add the topic / processes that can be logically sequenced]

Trial and error is key here. There are lots of prompts that can be used, and different AIs and chatbots will give outputs. Try, review, and work out which technology is best for your case.

A chatbot is really programmed to synthesise and summarise information. While it may appear “clever”, the bot is not a critical thinker. Nonetheless, AI tools like ChatGPT have potential to become an important tool for writers who want to hone their critical thinking skills along with their communication skills.

A trawl of some blogs and forums where researchers are engaging with ChatGPT reveals the following use-cases:

• “I use it to summarise articles so I can map out more efficiently what’s worth reading before I deep-dive into articles.”
• “Demystify complex text. Sometimes, there may be way too many new ideas and subjects introduced. Throw it all as a prompt and see what comes out.”
• “As a free tool, it seems good at shortening, lengthening, rephrasing, changing the tone.”
• “Use it as a guide or hint to subject matter information. Pretty much a Google 2.0.”
• “I fed ChatGPT some prompts and then asked it to generate an outline of the main chapter of my dissertation (already completed long ago).”

Looking for other opportunities and ways of using AI from blogs and forums, and making a shared class/group list of potential instructions, may be of benefit to you.

In the case of summarisation, it is advisable that you use the AI only for basic summarisation of text or a topic. Since the Chatbot is usually constrained by the word limit provided, it can miss out important points in a discussion, since it has no way to really summarise importance. This is illustrated by reference to the recent Turkey/Syria earthquake.

The following prompt was used:

List 5 reasons, in order of importance, as to why the 2023 Turkey earthquake killed so many people?

ChatGPT was unable to answer since it suggested the event was in the future! This is an edited response from Google Bard:

1. The earthquake was of a very high magnitude.
2. The earthquake struck a densely populated area.
3. The earthquake occurred in a region that is prone to earthquakes.
4. The earthquake was followed by aftershocks.
5. The earthquake occurred during a time of political and economic instability with poor building codes.

The answers are certainly reasonable, but a geographer would likely put them in a different order, starting with reason 5. Reason 3 seems somewhat irrelevant. The AI typically gives a generic answer, and in this case it has written about why big earthquakes often kill lots of people. This is one example of how the output looks convincing, but it is also possible to provide some deeper and more challenging instructions, e.g. “act as a statistician”.

Using ChatGPT, the instruction could be written as follows:

I want to act as a statistician. I will provide you with details related to statistics. You should have knowledge of statistics terminology, statistical distributions, confidence interval, probability, hypothesis testing, and statistical charts. My first request is “I need help calculating how many people in Manchester regularly visit green spaces”.

The response provided is shown below:

Try to test this statistical function with some of your own suggestions.

AI can also summarise lengthy information from the internet that is in a published form. For example, the following prompt could be used in Microsoft Bing with the “chat function” (it’s actually running ChatGPT):

TLDR: [followed by the link to your article or research paper]

A chatbot is not going to be able to write your NEA, and nor should you want it too. That’s because it can’t do your fieldwork or analyse graphical/cartographical information.

You should also be aware that, in all aspects of your NEA, independence must be maintained. If you use AI, then that needs to be acknowledged. The misuse of AI tools will be considered malpractice (JCQ, March 2023).

But there are some supporting activities that a chatbot may just be useful with – linked to different parts of the enquiry process. These include:

• Supporting you with the framing, development, and refinement of individual research questions, titles, or even hypotheses.
• As an efficient search tool to assist in narrowing down secondary information.
• Creating the literature review.
• Reviewing your writing style and produce the NEA in a more “academic” style.
• Shortening text.

In terms of supporting the framing of research questions around a possible topic area then this input would work well:

Develop three possible research questions on the following topic: “The influence of social media on place identity and representation”.

The AI responded:

1. How does social media shape the formation and expression of place identity in urban environments, and what are the implications for local communities?
2. What role does social media play in representing and promoting a sense of place in different geographical locations, and how does this influence perceptions and experiences of those places?
3. To what extent does social media contribute to the commodification and commercialization of place identity, and what are the consequences for local cultures and authenticity?

The AI was then prompted again:

Based on the second research question, develop three further research questions on the topic of social media influence on perceptions and experiences:

• Build my scientific method, qualitative discourse analysis, into each question.
• Include the term “place representation” in each question.

The AI then refined its initial output:

1. How do social media platforms contribute to the construction and dissemination of place representations, and what are the underlying discourses and narratives that shape these representations?
2. In what ways does the interaction between social media users and place representations influence the formation of perceptions and experiences of specific locations, and how do individuals negotiate their own identities within these mediated environments?
3. How do different user-generated content on social media platforms, such as photos, reviews, and narratives, contribute to the construction and negotiation of place representations, and how do these representations shape the collective understanding and meaning of specific locations?

Academic writing is a formal style of writing used at university. It’s something that your NEA should try to utilise as well. This is important for academic writing because it will prevent you from repeating the same words over again, and make your writing less personal, as the focus will be on an action rather than who did it.

Some examples of AI instructions/prompts to improve the quality of written work include:

There are some serious considerations to think about in relation to using AI. In particular, AI must be used as an assistive tool, not as a primary source of writing. AI language models do not know, think, or write like humans. They are, in many people’s eyes, “prediction engines”, or, at best, glorified encyclopaediae. Again, they do not in any way understand what they are producing / writing.

Some cautionary considerations of AI:

• The chatbot answers may look good, but are often inaccurate or unreasonable. The AI is trained on a large database consisting of blogs, Wikipedia entries, and webpages.
• Chatbots don’t directly reveal sources of information (although you can ask for, for example, Harvard-style referencing). This creates a problem for citing other authors' work directly.
• Chatbots are very sensitive to the question/problem posed. Slight changes in the input phrase may give radically different answers to what was (seemingly) a very similar question.
• Responses can be out of date, as the chatbot is basing its answer on internet information that is out-of-date or has been superseded.
• The output is often very wordy and overuses specific terms, words, or phrases, while at the same time lacking specifics.
• Due to the limited number of responses offered by the chatbot, search engines and published sources are still a reliable tool for research, and should not be substituted or replaced.

What do you think are the ethical considerations of using Chatbots with the NEA?

This author would argue that when used in the right way, chatbots can be a great tool to support learning rather than something to be feared. In the context of an NEA, technology can be used to augment and supplement some aspects of research, question development, and synthesis of findings. It’s important to stress, however, that a proper research and literature review should always be carried out, and you should overly rely on the information produced by a chatbot. The technology helps with writing edits, reviews, and giving you a “second pair of eyes”.

In the end, you must adhere to the usual rules around writing and the production of reports. Never pretend something is your own work if it is not, and always acknowledge the role of support in the form of a spell-checker or AI in the development of a piece of work.

Azaria, A. (2022, December) ChatGPT Usage and Limitations

Caulfield, J. & Solis, T. (2023, May) Using ChatGPT for assignments

f/awesome-chatgpt-prompts (github.com) (2023, July) Awesome ChatGPT Prompts

JCQ (2023, March) AI Use in Assessments: Protecting the Integrity of Qualifications

@Keenious (medium.com) (2023, February) ChatGPT in the Academic World

profkipp (youtube.com) (2022, December) Using ChatGPT for writing in research

Science Grad School (youtube.com) (2022, December) ChatGPT for Scientific Research...

The views expressed in this article are from the author and have not been endorsed by any awarding bodies. David Holmes works as a geography education and assessment consultant. He is also an author with a keen interest in fieldwork, GIS and the use of digital visualisation technology in geography. He can be contacted on [email protected], or visit www.david-holmes-geography.co.uk.