Chat GPT: How the controversial platform can inspire you in your PhD research and teaching, without breaking the rules.
‘Cheating’, ‘a broken education system’, or ‘unregulated dangerous AI’ might be the phrases you’d use to discuss Chat GPT’s recent emergence, but notice the software’s potential to be harnessed in academic teaching and research, and perhaps you can see the less obvious benefits…
For reasons including the risk to fair assessment, and a perceived threat to critical thinking, the education sector is deemed to be frightfully unprepared to deal with the enormous paradigm shift that is the development of AI natural language processing software, otherwise known as large language model (LLM) software.
It appears true that if these technologies do not receive proactive regulation from governmental or non-governmental bodies, as their software is further refined, there could be greater threats to learning and assessment in secondary and higher education. Recent findings suggest that almost 50% of American university students use Chat GPT in their formal examinations, and suggest similar rates of use in Europe and around the globe. Additionally, the Technical University of Munich published findings that the software will cause overreliance from students leading to a loss of critical thinking, summarising, and verification ability, ultimately affecting students’ learned development.
Considering this, it may be surprising to know that the careful adoption of Chat GPT in teaching may directly mitigate some of these issues, and mindful use from researchers has incredible implications for time-saving, refined searching, formatting, and more.
Harnessing AI for Teaching in Higher Education
- Critique the content. PhD educators have already begun using AI to generate academic work and critique its content with their students. Chat GPT’s formatting and natural language mimicking capabilities can make their generated academic responses seem advanced and well-structured, yet in analysing these responses with your students (a dynamic teaching method) you could teach them to identify the limitations of these responses. General limitations include, for example, poor content, not using the required structure, missing details, no referencing, and a poor ability to integrate contextually specific theory into writing.
- MOOC courses. As a PhD teacher or ECR teacher, you’re most likely stretched for time. It may feel like taking a free online course is too time-consuming, yet brief courses on AI for teaching may grant you the expertise to employ this software and cut the time you spend producing teaching resources like lesson plan curation in half.
- Breaking barriers. Chat GPT demonstrates enormous capabilities to help graduates and postgraduates for whom English is a second language to clarify what they mean in essays and other written forms. Accessibility for non-native speakers has long been a problem in academia, but a study demonstrates that using LLM to rephrase what students mean in an academic register reduces the cognitive burden of translating, allowing for more equitable amount of time to be spent focusing on content akin to native speakers. Making your students aware of this simple feature, if within your institution’s academic integrity guidelines, could help many students increase their performance.
- Catering to learning needs. The education sector recognises the growing importance of modifying teaching delivery to accommodate students with learning difficulties. Chat GPT can efficiently produce audio and visual learning tools to support students with specific needs, for example, students with dyslexia or visual impairments. Even if you cannot use the interface in teaching, sharing this information could have real value for disadvantaged students.
Chatbots for Academic Research: Your best friend or your enemy?
- A search engine? Although not a search engine in computational terms, Chat-GPT is trained on 175 billion parameters, and can answer more contextual and nuanced questions about challenging content you come across in your research. Just make sure never to use Chat GPT’s text in your academic writing, it is often inaccurate, with poor fact-checking capabilities, and can in many institutions be traced as AI generated discourse.
- Summarising readings. Spending too long on your recommended readings? To avoid reading endlessly dense publications, with what some may call an ‘opaque writing style’, you can ask the processor to read a large text (up to 3000 words at a time) and summarise it into key findings, or to rewrite in simpler terms. Be aware that AI summarisations often exclude some of the authors’ arguments and can utterly misinterpret the text input.
- Proof-reading references. Spend hours on referencing? Although there is general scholarly acceptance that the software is poor at producing references, Chat GPT can reformat your references based on your preferred parameters including citation style, as well as check for simple referencing errors. This efficient process automates reference organisation, saving you valuable time and effort while maintaining accuracy and consistency.
- Find inspiration. If you’re having a methodological crisis, doubting your research’s relevance in your chosen field, or simply out of ideas while writing a paper, run your thoughts or questions past Chat GPT. The software will not produce your dream thesis or paper, but it is remarkably good at helping you take a different perspective on things or nudging you towards particular ways of seeing things that you may have overlooked in your own research. Try asking, for example, “what are some alternative ideas in EU security studies that challenge academic theory?” and then dive deeper into the topic through traditional resources.
Concluding: A researcher’s silver bullet?
This article exemplifies the breadth of opportunities LLM software has to offer academics as researchers and teachers, but also notes serious limitations. It is essential to be conscious of these limitations and be mindful of the effects that AI-generated content could have on your ability to critically engage with your research and be independent in your perspectives.
As of July 2023, Chat GPT has only processed data up to September 2021, and therefore is far from up to date with latest paradigms, publications, and current events relevant to your research and specialisms. For instance, geopolitics are dynamic and everchanging, yet the processor is quite oblivious to the last 2 years of the Russo-Ukrainian War, and therefore cannot be the most useful tool for researchers in contemporary EU studies. Even the software agrees! Try asking ‘are you up to date with current events?’.
Always verify with your institution if and how AI software can be used in your research or teaching and take the steps to learn how to use it effectively, without falling into the perilous trap of copy and paste.