Have you ever received confusing letters from the government or companies? Sometimes, it’s tough to understand what they’re saying. Way back in 1979, people recognized that these letters should be simple and easy to understand. Since then, efforts have been made to make them clearer, including creating rules for simple writing. But, creating change hasn’t been easy. With new ways of communicating, like digital methods, the problem has become even bigger. PNA Group asked Nadine Beks van Raaij to look at ways to create a solution.
Miscommunication is widespread. PNA Group posed a question: “How can we reduce miscommunication and bridge the gap between government and citizens?” By making the written communication simpler, the aim is to create an improvement.
Nadine’s study delved into testing three different models. She used both technical and human evaluations to figure out which model worked best. Nadine did technical evaluations and interviews with experts. When the first two models didn’t meet the mark, she shifted her focus to the GPT model. The GPTmodel proved most effective. This was followed by a reader research/randomized control trial involving 72 participants from different educational backgrounds. This ensured the results reflected society as a whole.
Participants were shown three letters. They were either shown the original letter or version simplified by the GPT model. Participants then answered questions to test how well they understood the information. The research aimed to assess understanding through questions like, “What’s the next step to help your aunt apply for a stairlift?” and “Do you understand the meaning of this word?” The findings showed that comprehension increased from around 60% for original letters to over 90% for the simplified versions. Nadine conducted regression analysis and examined correlations with factors like age to validate the results. It confirmed logical assumptions, such as older people spending fewer hours reading each day for their work as they are retired. This ensured her research wasn’t influenced by unexpected factors.
The tool Nadine developed doesn’t replace human involvement but assists it. People remain responsible and in control. Prompt engineering, is a new approach, that aims to optimize outcomes by asking questions. Nadine is now designing an interface with embedded prompts to ensure consistency for users, especially those working at government agencies.
This study, by Nadine Beks van Raaij, sheds light on how machine learning, particularly the GPT model, can significantly enhance government communication by reducing miscommunication. This not only improves comprehension but empowers citizens to make better decisions. By simplifying complex information, we bridge the gap between government organizations and citizens, ultimately contributing to a more informed and engaged society.
“We are very pleased with the results of Nadine’s work – not only are they useful for current implementation, but they are also a promise for exciting further research. Nadine is a joy to work with – a bright, open and creative person, with the right mindset and excellent communication skills.” – Frank Harmsen (Managing Director PNA)
We sat down with Nadine Beks van Raaij recently graduated Cum Laude from the Master’s program in Data Science for Business and Entrepreneurship (DSBE). Her passion for making a real impact through data science led her to ace her master thesis. In this interview, Nadine shares her journey, her thesis work, her experiences at JADS and what’s next.
JADS provides innovative data science programs at undergraduate, graduate and postgraduate level, conducts cutting edge data science research and offers great business opportunities with a continually growing ecosystem. From startups to SMEs and large corporates, our business partners gain access to talent, high level knowledge and business opportunities.
Partner up with JADS like 300+ other organizations and businesses and shape your data-driven future.