Enhancing proposal writing with ChatGPT 3.5
Streamlining the process to save time for researchers and grant writers alike
I used ChatGPT to provide the title for this article. I stared at my article for 15 minutes, didn’t come up with a title and used ChatGPT. It saved me time and I could prompt it a few times to try again, so that I had multiple choices.
ChatGPT is a language model, it is designed to predict the most probable next word, and that is what it does, no more and no less. It won’t take over the world, it cannot write your research papers or proposals by itself, it is not a self-thinking, all-knowing entity. It is a language tool. And if you treat it as such and don’t expect more, and be careful with not asking for more (don’t ask for a list of references for example, it is a language tool, it will give you what fits language wise and not necessarily what really exists), then you will be able to use this tool effectively, save time, some frustration and maybe even learn a thing or two about writing and storytelling.
Now that we established what to expect and what not to expect, let’s go into the use of ChatGPT for proposal writing. We tried it out on an awarded proposal from the Horizon2020 program, so most of our tips are based on the sections of these templates. European Horizon2020 proposals consists of an:
· excellence part describing the challenge, state-of-the-art, ambition and methods of the project
· impact part describing the expected impact of the project and asking for a dissemination, communication and exploitation plan
· implementation part describing the work packages and management process for the project
We tried ChatGPT mainly for the excellence and impact part, but can also see potential use for the implementation part.
Before we get started, two disclaimers from my side:
1. It doesn’t matter for which section or text you use ChatGPT, keep in mind that these texts are drafts and should ALWAYS be reviewed and/or adjusted.
2. NEVER feed ChatGPT sensitive data, such as patient information. It is an OpenAI tool, which means everything we put in can potentially be exploited by OpenAI.
And now to the practical part. We tested Chat GPT3.5 on an awarded Horizon2020 project for which we received early, middle and final drafts and could compare ChatGPT outputs with the actual outputs of the final proposal. The language tool was very useful doing the following:
To shorten or summarize a text: this is especially useful for the abstract and work packages or any section where you need to meet the word count. You wouldn’t believe how much time researchers and grant writers spend on shortening or summarizing text. Here, ChatGPT is extremely useful and relatively reliable and can save many hours. As always, check the text afterwards to make sure nothing crucial was left out, but I bet you will astonished how well that works.
To rewrite a text: ChatGPT is amazing in rewriting text, but we were astonished how well it did in rewriting text given specific prompts, as for example reviewer comments or better addressing the guidelines provided by the funder for this section. With a specific command such as ‘extend this text to 500 words and rewrite incorporating this reviewer comment’ we received very good results.
To translate a text into a table: this is a useful feature for dissemination and communication plans, just provide ChatGPT with the text, and a clear command in which table set-up you need this text, e.g. how many columns, title of columns etc. and check the results. You might need to adjust here and there, but the general outcome was good and saved time (and also provided some extra ideas, since ChatGPT fills in text gaps if they occur, so if you have 3 activities in 3 lines and a stakeholder column for each line, but you only provide stakeholders for activity one and two, ChatGPT will add stakeholders for activity 3. Often enough, they are accurate, but always check).
To translate bullet points into a coherent text: in first draft of proposals, we often have sections where the outcomes of a brainstorm were hastily wrote down in bullet points. This is something that ChatGPT can write out into a good and coherent text. It saves tons of time and only needs to be reviewed and extended in next versions.
Funnily enough the other way around didn’t work so well, translating a text into bullet points, unless…
To use specific writing or planning tools on a text: …you used a specific tool, such as SMART goals to translate your text into bullet points and then make them SMART or FAIR.
SMART goals: tool used to plan and achieve goals, making them Specific, Measurable, Achievable, Realistic, Time-bound.
FAIR principle: a useful framework to optimise the reuse of data. FAIR stands for Findable, Accessible, Interoperable, Reusable.
Using ChatGPT to rewrite your text into SMART goals or FAIR principles is especially useful for impact sections, to address expected impacts and give a data management plan that follow the FAIR principle. But be aware, in case you didn’t provide all the necessary information in your source text, ChatGPT will again include text to fill gaps, which can be a great prompt for ideas, but it can also create wrong outcomes, so please review these areas thoroughly.
And the final tip how ChatGPT can be useful is simply to help you if you have the blank-page-syndrome. We have all been there, especially on a Friday afternoon, which is by the way also the perfect time to play with ChatGPT, it will energize you for the weekend.
If you have questions please leave them in the chat and I will try to get back to you as soon as possible (and with the limited knowledge I have. I cannot answer tech-questions about ChatGPT, for these you will have to ask ChatGPT yourself 😊).
