Using AI tools to practise summarising

 In the UK, apart from Scotland, summary of an aural and a written text is part of the advanced level exam (GCSE A-level) taken by A-Level languages students. It is arguably one of the more challenging tasks to be done in exam conditions and requires a good deal of practice during the course. the task requires students to get across key comprehension points, whie manipulating and altering the language of the source text. This is particularly hard with the audio text. 

In general, it's a skill which can be gradually built over the course, just like the film and literature essays, through 'micro-skill' exercises which contribute to the final 'macro-skill'. Much better to do this, of course, than to throw students into the deep end by just giving them a summary to do with no preparation.

For example, students can do exercises such as:

• Paraphrasing single sentences.
• Doing morphology exercises (relating nouns to verbs, verbs to nouns, adjectives to adverbs, etc).
• Asking students to paraphrase responses during oral work.
• Doing gap-fill tasks with various options to choose from for each gap.
• Doing certain structural drills, e.g. active to passive, passive to active.
• Doing sentence combining exercises, or the opposite, making two senetnces out of one longer one.
• Modelling the paraphrase of words, chunks, sentencing and whole extracts (using a 'think aloud' approach).

AI tools can come in useful when designing some of the above the tasks, but clearly a significant issue is the risk that students will resort to using them to paraphrase whole texts without doing the mental processng themselves. Many students (in my experience) would be honest, but the temptation to cheat is there. As a result, some teachers will choose to avoid setting summaries for homework. This is a pity, since I believe class work should be mainly devoted to listening, reading and speaking.

So how can we use an AI tool to positive effect?

1. How about taking a source text, and asking students to produce a paraphrase using ChatGPT? They just need to copy in a text after giving the instruction: Summarise the following text in French, in French, using the same number of words (or a particular word limit). Students should then compare to the two texts and write notes on the differences, e.g. synonyms used, new structures, new vocabulary, possible inaccuracies. By doing this, they will develop their paraphrasing metacognitive skills (thinking about and monitoring how it's done).

Students take a source text and produce two different paraphrases using two different AI tools, e.g. ChatGPT and Bing Chat. In note form, they analyse the differences between the two different versions.

2. At the word or chunk level, have students (in or outside class) find synonyms, by using an instruction such as Give me a synonym for the following French word: problème. They can then compare the results in class and insert the words into sentences.

3. To improve students' skill in extracting the key comprehension points froma text, have an AI tool produce bullet points from a text.

In the example below, I asked ChatGPT to produce 10 bullet point notes based on the text below (about a French government initiative to redice bullying in schools).

Le gouvernement va revoir les programmes scolaires. Le ministre de l'Education nationale, Gabriel Attal, a annoncé en septembre 2023 la mise en place de « cours d'empathie inscrits dans le programme scolaire », « à l'image de ce qui existe dans d'autres pays, notamment au Danemark », qui seraient généralisés à l'école à partir de la rentrée 2024. Dès janvier 2024, une école pilote dans chaque département expérimenterait ces cours.

En 2023 on a assisté en France à un « tsunami de témoignages venant des établissements scolaires. Ils ont été multipliés par trois par rapport aux chiffres de septembre 2022.

Plusieurs mesures avaient déjà été mises en place en 2023 pour renforcer la lutte contre ce fléau, dont la possibilité de changer d'établissements les élèves harceleurs. Cela évite d'imposer ce changement à la victime, ou de pouvoir sanctionner un auteur de cyberharcèlement contre un élève d'un autre établissement. Le ministre Gabriel Attal a affirmé vouloir une « tolérance zéro contre toutes les formes de harcèlement ».

Le harcèlement à l'école touche un élève sur dix, selon les associations. En septembre 2023 il était au cœur de l'actualité après un nouveau drame, le suicide du jeune Nicolas, 15 ans, à Poissy (Yvelines) le 5 septembre.

Here are the bullet points ChatGPT produced:

1. The government will revise school programs in France.
2. Minister of National Education, Gabriel Attal, announced in September 2023 the introduction of "empathy courses integrated into the school curriculum."
3. These courses, similar to those in other countries, especially Denmark, will be implemented nationwide starting from the 2024 school year.
4. In January 2024, a pilot school in each department will experiment with these courses.
5. In 2023, France experienced a significant increase in reports of bullying and harassment in schools, with numbers tripling compared to September 2022.
6. Several measures were already introduced in 2023 to combat this issue, including allowing bullying students to change schools.
7. This change aims to avoid forcing the victim to switch schools and enables sanctions against cyberbullying perpetrators from different schools.
8. Minister Gabriel Attal emphasized a "zero tolerance" approach to all forms of harassment.
9. According to associations, school bullying affects one in ten students.
10. The issue gained prominence in September 2023 following a tragic incident, the suicide of 15-year-old Nicolas in Poissy (Yvelines) on September 5th.

Students could then translate these sentences into French.