Day 2 – Summarising text
Welcome to Day 2 of the 12 Days of AI. Today, we will be looking at a AI tool called Claude which is similar to ChatGPT. Claude allows users to upload PDFs and Docs as prompts and extracts text to generate summaries.
What is it?
Claude is a large language model, similar to ChatGPT and it was created by the AI company Anthropic, an AI startup based in San Francisco. It was first released in July 2023. Claude utilizes a conversational model trained on Anthropic’s dataset, allowing it to engage in open-ended conversations and claims that it can avoid harmful, unethical, or illegal language. Claude’s privacy policy.
How do I access it?
First you need to create an account with Claude.
Your task:
Upload a Word document or pdf to Claude and ask it to summarise the text for you. Try out one of these tasks:
Task 1:
Claude could be useful if you want a quick summary of an academic article, report or if you are conducting a literature review.
- Select the paperclip icon and upload a pdf document.
- Ask Claude to summarise the pdf for you, in a limited number of words, including a title and sub-heading.
Task 2:
Claude can do a similar job summarising Word documents too. Again, this could be useful for summarising some everyday tasks that could be quite time consuming. For example, you could use it to summarise the contents or the individual contribution of someone in a online seminar or supervision meeting.
Try downloading the transcript from a MS Teams (or Zoom) meeting and then upload it to Claude and ask it to create a short summary for you. (Make sure that you have permission to do so from the other meeting participants.)
- Upload a Word file using the paperclip icon and add a prompt asking Claude to summarise the document.
Optional reading:
Assessing AI-based Summarizing and Paraphrasing Tools for a CLIL Intercultural Communication Academic Writing Class by Barry Kavanagh (Associate Professor at Tohoku University, Japan)
Join the conversation:
Reply in the comments below or post on X using #12DoAI.
- How did you get on with the task – did you find it useful?
- Summarising and paraphrasing text is a key writing skill that we encourage our students to develop – does this use of AI undermine this skill?
- Should we be using it and promoting it with our students?
- Do you have any privacy concerns using Claude (or ChatGPT)?
- The quality of the output generated by these tools can be questioned and can create problems rather than solve them. What did you think of Claude’s summaries –were they any good? How else could Claude be useful to us in our professional practice?
Join the competition
We’re running a competition during this course where you can win actual prizes! Learn more aboutt how to enter the competition.
I asked Claude to produce 10 questions on the Tudors, of these 2 were completely wrong and another 2 had missed some details so not a great success rate, although it was very polite when I corrected it!
I also asked it if uploaded documents are stored, I’ve copied the response below .
No, documents are not stored when uploaded to Claude. Claude’s functionality is based on processing and summarizing text in real-time through natural language processing. Any documents or text passages shared with Claude are analyzed in the moment to generate responses or summaries, but not retained or saved after the conversation ends. Unlike some other AI systems designed for long-term personalization based on a user’s data, Claude does not have persistent memory or user profiles. Each conversation starts fresh. So you can feel comfortable uploading text to Claude without worrying about privacy concerns related to data retention. Any documents are fully anonymous and transient rather than stored by the system.
Thanks for asking this question, Susan. This is reassuring.
Thank you for sharing this, Susan! It’s a very important question to ask as we consider what content we share with these AI tools.
It took the comments from Day One of the 12 Days of AI and copied them into a Word document and then asked Claude to summarise them for me. I then put the summary into ChatGPT and asked it to create a blog for me. This is what I got:
https://totallyrewired.wordpress.com/2023/12/01/how-chatgpt-is-revolutionizing-education-a-detailed-analysis/
that’s impressive matey! To create the blog did you use ChatGPT 3.3 (free) or 4 (paid)?
I used 4 but I’m sure 3 can do it too!
I asked Claude to summaries key points of a Thesis on Kinetic garment construction without word count and it gave a fair overall view. I then asked Claude to identify historic/people references and clearly spotted that the theses is an account of the author experience in different fashion houses which focused on outlining conceptual frameworks rather than including direct quotes or insights from interviews. Useful tools. I try to create a blog with a written doc on Chat GPT 3.5 but not success.
Hi Christopher can you please be more specific on how you created the blog with Chat GPT. Thanks
Hi Nadia, I have created several blog post from webinar discussions i’ve organised. First, I downloaded the transcript form MS Teams. Next, I up loaded the transcript Word document into Claude and asked it to summarise the document in 900 words. Finally, I wrote in prompt in ChatGPT saying ‘Write a blog post of 500 words on the following information: and pasted Claude’s summary into the prompt.
hope that’s clear 🙂
So I took your blog post, pasted it into Word and put it through Turnitin… 20% AI. They say anything below 20% is more likely to be a false positive anyway.
very interesting Steve ….
The blog post was a genius idea!
Dont know about ‘genius’ but it was easy to do!
I found the Claude app useful! As someone who needs to break down large passages of text into smaller doses or summaries to gain understanding, something like this would’ve been incredibly helpful during my years of study. It wrapped everything up very succinctly and got straight to the point.
Although useful, it *may* have the potential to become relied upon or used as a first resort for students, and over time, impact their summary and paraphrasing skills, as a result of consistent automation and over-use.
I think if used as a supportive aid rather than the action taker, it could be a really useful tool for students, especially for those who have processing difficulties. I think there would perhaps have to be a conversation around exercising discernment and restraint with the use of such tools, with total transparency in regards to how over-use can/may affect skill-sets in this area. A nod to the whole “if you don’t use it…” saying!
I agree with you, Kei. I was impressed by Claude’s output – I asked it to summarise a text in French and it did a good job.
Like you, I am concerned that the skills of summarising, reformulating, etc – which are key in language teaching and bring so much to the students – will be lost. They would be missing out on the whole process, which in my view is enormously helpful to them.
In a broader sense, I do worry about everything ‘speeding up’ with these tools. Will we have time to stop and think, evaluate, reflect?
I uploaded and PDF and then a Word document and got a reasonable summary of the content in key points. Surprised by how efficient and quick it was.Because it doesnt retain the content that satisfied the data protection issue.
I asked Claude to summarise a PDF for me, but encountered a copyright issue. Instead, I asked Claude to summarise my own essay and it was able to give a surface level description of the key points and sections. I think I would find this method a bit difficult to use consistently, but it would probably be great for making sure your key points in a thesis are distinguishable enough.
I agree with the comments above, it’s potentially great if you use it carefully, especially for inclusivity. I guess it would be useful to know how consistent a summary was and you’d need to be careful that you don’t miss out on the level of exploration and understanding that is possible when you read something fully, or at least skim and delve into sections in more detail. As ever, a useful tool, but not the whole answer (especially if it’s wrong!).
I’m really interested in the comment above about how it doesn’t store the documents, etc.. That’s potentially really reassuring, but I wonder how this will be monitored.
I guess the fear (is it fear) is that it will be addictive or totally relied on for certain people, students of course maybe the prime example. But is this a bad thing? It’s like the way we rely on “predictive text’s’ or Word corrections, or indeed Grammar AI software. It is just another stage I guess, and although it has it’s (potential) issues, used in the correct way it can also be stimulating and refreshing! I find it useful to summarise papers and large documents. It is potentially a new way of learning for a new generation…
We still need to think!
Useful to obtaining the key details and themes if you have a large amount of reading to do: I asked it to summarise a long Word document that I hadn’t read and it produced an articulate and easy to read summary. I can see how this could speed up the rate at which you could read a collection of papers, but perhaps you would miss the detail, and I wonder what it might miss?
Not so useful if you are summarising/paraphrasing for learning: The practice of summarising and paraphrasing is a good exercise for learning; it forces you to explain the ideas in the text and thereby you explore your own understanding of it. I don’t think having a machine do this for you is useful in that scenario.
Hi Anna, this is such an interesting comment and made me reflect on the ability to summarise text as a skill. Claude will do it for you, but it would be interesting to understand how it summarises and how it edits before using it. Maybe we could use it as a teaching tool where we ask students to make a summary of an article and then compare it with one created by Claude. It might spark some nice reflections.
I have used it with subject experts, where I have asked Claude to summarize a transcript of a recorded lecture that they had delivered. They can tell weather or not the summary is accurate and/or incomplete and use it as an additional resource for the student. But your main key point still stands; the lecturers would have known how to do it even without Claude, it would have just taken them longer.
I agree summarising and paraphrasing are good acaemdic skills or learning. My daughter is in her final year of a UG degree told me she uses AI to summarise articles all the time to see whether it’s worth reading, or to home in on the part that is relevant that she wants to read more about. So in this example it’s not a replacement for that, it’s about saving time for her so she’s not reading lots of articles that aren’t useful/relevant.
I agree with Anna, I think Claude might be useful to give short summaries of a range of research papers, perhaps allowing students to pinpoint the ones they want to drill down into in a bit more detail. The concern is that important detail might be missed.
Students will use it, I am sure. The challenge is getting them to understand the learning they are missing by over reliance on the tools.
The advantage of using it as a summariser is that by uploading a pdf of a research paper, students can be clear where the information has come from and properly attribute it.
I submitted one of my academic papers and it gave me back a really good summary!
Hi used Claude to upload a job description. I asked it to identify key action verbs along with skills, competencies and behaviours mentioned.
The output on the screen was really useful as it categorised the areas neatly and allowed the document to be more easily digested.
However, unfortunately, it missed out some key competencies that were ‘inside’ longer sentences and required an element of nuance and interpretation.
I felt the task I’d given it required more than simply a summary and therefore, it didn’t seem to ‘cope’ as well with that.
Loved doing this task! 🙂
Todays task was really enlightening for me! I have been using an AI document bot called ChatPDF ‘ (by pure coincidence) from some testing I carried out yesterday in which I uploaded a software ‘User guide it was regularly giving me incorrect answers (even when I told it that it was wrong the follow-up ‘corrected’ answers were also wrong). I tried uploading the same User guide to Claude and asked it exactly the same questions and it gave the correct answers first time! I am now a Claude convert!
I uploaded a transcript from a 1 hr online lecture which was 9000 words. Asked to summarise in to 1000 word it did a reasonable job. The speaker referenced some theory, and when asked, Claude gave 5 real references to papers not mentioned. Quite useful to add to my written notes of the talk. But, the lecture was inspirational with rich insights, none of which were captured in the summary. That’s probably asking too much.
I would have liked to drill down further to discuss some topics mentioned in the talk but quickly hit the free message limit on the account. Having an ethical position is interesting, but I think the details on that are key – I mean they still use prompt and output date from conversations to train the model, just in a more considered way re privacy. According to Wiki Anthropic are receiving USD 6B from google and amazon.
Great tool for specific activities. A bit cautious about students using these tools without any criticality. Its use should be scaffold for students to gain understanding and practice. Learning needs to be active. If student engage with the tool and the material, the better the learning. Need to maintain the process of learning rather than promoting the final product.
I uploaded a pdf journal article and asked for a summary in 200 words, after checking through the article myself, decided it did a fairly good job user friendly language. I would of liked it to acknowledge/mention some of the research behind the findings.
In terms of privacy not retaining documents/pdf uploaded that is a good thing, however why was it necessary for me to submit my mobile number to register, is that deleted?
I agree quite strongly with Anna and others in respect of ‘we still need to think’ this is crucial to understanding the topic you are reading about, if learners relied on this as a crutch to get by and pass assignments, in reality they are getting certificates to get them into jobs they are not qualified (lack knowledge) for where they can do more harm than good. Instilling in learners how to use these tools ethically is paramount.
On day one I tried Claude as well as ChatGPT and found the response from Claude not as thorough and in depth.I also chose Perplexity which gave a misguiding response to exactly the same prompt.
For todays task my first report (in Claude) was rejected because it was too large, second report Claude said it could not properly analyse but the third (one of my lectures from last year) brought an impressive summary!(pleased it was actually quite a good lecture then…)
ChatGPT had no problem with either of the first two reports and gave a more detailed response to the third.
I found it will be useful to get the gist of a Teams meeting that I would’ve liked to attend but couldn’t.
I did struggle with attaching the document and adding the prompt in the right order.
I asked Claude to provide a 200-word summary of a 13-page academic article by John Field on developing students’ processing skills while listening. I think it provided quite a good overview of the main and supporting points and offered some critical evaluation of the limitations of certain approaches to listening.
I feel that if you set a rather low word limit to the summary it will be an effective tool as it provides an overview and structure of the text but the students will still need to fill in the details and apply critical thinking to evaluate the key points of the article, which means that they will still need to refer to or read the article. If you set a high word limit, then the need for such exploratory work may not be needed.
I used Claude like a feedback tool, which was really handy, based on a guide I had created myself.
– I started by asking it to summarise the guide into its top 10 points
– I asked it which point it thought was most important and why (and was pleasantly surprised by its response)
– I asked it how the list could be improved and much like a tool like Grammarly, it suggested more storytelling and motivational tones – but where Grammarly just supports you with Grammar tones, I’m impressed at how this tool can quickly support you to strike the right tone across the entire document and give examples of ways of doing that – probably not so great for students who are being marked on things related to doing this independently,but for people who are working in roles where it would be great to claw back more time in your work day (working smarter, not harder), this is great
– I finished by asking it if it could spot any ethical issues related to the content of the document and it couldn’t, but still it helpfully went on to reflect on bias, overgeneralisation, and prescriptiveness, to allow these things to be considered when making any further edits.
Just created this blog post using Claude. It is based on the ninth AI Conversation session that took place today where Emma Gibson answered questions around the topic, Gender Bias And AI – Can We Do Anything About It. The session was recorded in MS Teams, which automatically created a transcript. The transcript was summarised in Claude and the blog post was generated using ChatGPT4. Image created by DALL-E 4.:
https://totallyrewired.wordpress.com/2023/12/01/tackling-gender-bias-in-ai-insights-from-emma-gibson-and-audri/
It’s distracting to have the process up front in the blog. I found myself skimming the article and wondering if it was fake?. Perhaps it is missing the information that a human was in the meeting and a human used the genAI tools and (hopefully) a human verified the text against the original event.
I’m swirling in issues of trust
As AI rain pours down
To unsettle the digital dust
Whatever ‘I’ do, dancing with robots
Reflection and testing is a must.
#12DoAI poetry)
I asked Claude to summarise a 4500 word creative text about a complex set of intertwined conspiracies, written in a bizarre stream of consciousness style. Claude did a great job and presented a very reasonable 300 word summary. Claude was also able to pitch the summary at younger readers and offer some clear explanations for the strangeness of the original text.
its an interesting tool and i can definitely see how it could be useful to students and ourselves however we hit straight away a copyright issue, unless you own the copyright in what you submit as a pdf then you cannot legally upload it to a site such as Claude…i also asked Claude this and their reply…..
“No, uploading full-text copies of academic articles to Claude that are written by others and published would likely infringe their copyright protections. Here are some key points on this:
Academic articles are the intellectual property and creative work product of the authors and the publishers that disseminate them. As such, they enjoy protections under copyright law. This generally gives the copyright holders control over and the exclusive rights to their reproduction and distribution.
Uploading full copies of articles to Claude without permission from the rights holders would thus normally constitute copyright infringement. There could potentially be exceptions under fair use doctrines, but systematically reproducing and providing full access would likely exceed typical fair use provisions.
Many open access licensing schemes exist to allow legal sharing and dissemination of academic works. However, restrictions would still apply in most cases against unauthorized full-text uploads to an AI system like Claude.
In summary, no, currently Claude does not have legal permissions in place for users to freely upload full-text academic articles written by other researchers or entities into my knowledge base. Any articles my responses reference or quotes come from have been accessed through proper scholarly channels and databases. I’m an AI assistant built by Anthropic to be helpful, harmless, and honest.”
but of course people will and students definitely will be uploading articles to Claude
You are absolutely right to flag copyright issues. I think this needs to be considered across all tools. The criticisms of many AI tools is – where does it get it’s info? And who inputs it?
Very interesting Claude’s AI used the phrase “likely infringe”… wonder what it would take to be “definitely”?
i think it might be to do with the licence of the article, as some of them will have a Creative Commons licence which could allow this to be added and some will not (the majority)
That’s interesting, especially given that uploaded files are supposedly transient and held in ‘working memory’, rather than added to the model. The answer from Claude above suggests these texts are added to the knowledge base. For my research I have been using HumataAI for summarising adademic papers, and it has been very helpful. However like most of these platforms you need to subscribe to get real use. Having cancelled all my digital media subscribtions I find I’m spending even more on AI tools !
I am a very old-fashioned learner and teacher. I am so grateful and happy to know the AI tools in research! I am totally amazed by the 2 tools introduced during the last 2 days! Also so useful to read the discussions from others. Claude is more easy to use with long files compare to ChatGPT, but quickly hit the free message limit on the account. Cant wait to learn more about other tools!
Thanks for joining in! I think the trick with these tools is finding out which ones work for you and your needs. Hopefully you find some that help you blend the old and the new styles of learning and teaching.
I uploaded a journal article to Claude to generate a summary within 200 words. The summary does capture the gist of the paper. I agree with some of the viewpoints here about over-reliance on summaries. As I have already read the paper and identified what’s relevant or important to me, I found the summary quite generic. If a student just read a summary of a paper without even skimming through it, they may miss quite a lot of interesting or important ideas, data and arguments in it. Perhaps, an AI generated summary would be more useful when the purpose of reading a text is just to get the gist of it.
I also asked Claude to paraphrase one sentence in the article. The paraphrase was actually quite good. The AI broken down the original sentence, which is very long and complex. By coming up with several short sentences, the AI has certainly changed the original grammatical structure, which in my view is the most effective technique to do paraphrasing. The AI also used some synonyms quite effectively. However, the adjective ‘scholarly’ added before ‘literature’ in the paraphrase sounds redundant. Furthermore, as the AI only focused on the sentence, there is no way for other information from the rest of this article to be integrated into the paraphrase. All in all, the paraphrase outcome was good, certainly with some room to improve.
I would argue that the summarising and paraphrasing functions are very useful to students, especially EAL students who are struggled with English language. However, this does not prevent tutors from teaching or guiding students how to do summarising and paraphrasing. With the skill, students could utilise such an AI tool more effectively and critically.
I found it interesting to discover another AI tool, Claude, which allows users to upload entire articles. I was impressed, as usual with large language models, by the speed at which it generated responses. Claude’s feature of attaching a file makes tasks even easier compared to ChatGPT, which requires copying and pasting content. The quality of the summary for the specific article I uploaded seemed similar and might help gauge whether reading the entire piece is worthwhile. However, solely scanning through, our usual approach, might yield a similar or perhaps a more reliable result.
The worry that students might skip reading and summarising, potentially harming their learning by focusing on the end result, is concerning. As it stands, I wouldn’t recommend this tool to students, but they might discover it independently, gaining an edge compared to others. Privacy is also a concern, especially when apps request email and telephone number details. Finally, I’m apprehensive about potential copyright issues; using such tools might lead to storing, sharing, or utilising our research without consent.
I asked Claude if it can use British spelling instead of American when paraphrasing texts. Here’s the reply:
Yes, I can use British spelling instead of American spelling when paraphrasing or writing texts. As an AI assistant created by Anthropic to be helpful, harmless, and honest, I am capable of changing my default settings to use British spellings and vocabulary suitable for UK contexts when needed.
Some of the common spelling differences I keep in mind include:
“ize” word endings become “ise” (e.g. recognize becomes recognise, organization becomes organisation)
“or” word endings become “our” (e.g. color becomes colour, flavor becomes flavour)
Doubled consonants in certain words (e.g. canceled becomes cancelled, traveler becomes traveller)
Changing “defense” to “defence”, “offense” to “offence”
Changing “curb” to “kerb”, “accessorize” to “accessorise”
In terms of vocabulary and expressions, I can use phrases like “whilst” instead of “while”, “roundabout” instead of “traffic circle”, “autumn” instead of “fall”, “lift” rather than “elevator” or other commonly used British terminology when paraphrasing, summarising or in any written text.
As an AI without a fixed region, I’m flexible to adjust my English spellings and conventions to your preferences – so just say the word and I will ensure my paraphrasing uses Australian/British standards for any assignments or writing projects.
Claude did a good summary of some lecture notes for me an picked up the key points. I asked for some MCQ questions – SBA – they came a bit more slowly than ChatGPT yesterday. One of the answers was wrong which I challenged and Claude apologised!! Unfortunately I then ran out of free time – which is a shame.
I asked Claude and ChatGPT about my institution’s response to AI.
Claude’s response: Unfortunately, I do not have enough context to summarize Xxx University’s response to generative AI. As an AI assistant without direct access to source documents, I would need more details or original text from Xxx University referring to their response in order to summarize it accurately. Please provide the relevant original text or documents, and I would be happy to analyze and summarize them to the best of my ability within approximately 300 words. Let me know if you can provide any additional context that would help me create this summary.
ChatGPT;
As of my last knowledge update in January 2022, I don’t have specific information about Xxx University’s response to generative AI The information available may have changed since then. Therefore, it’s essential to check the latest official statements or communications from the university for the most accurate and up-to-date.
In general, when educational institutions respond to generative AI, they may consider several factors. Generative AI, which includes technologies like GPT-3, can automate content generation and pose challenges related to academic integrity. Educational institutions typically respond by incorporating guidelines, policies, and technologies to maintain academic standards and ethical practices.
Some common elements in a university’s response to generative AI might include:
Academic Integrity Policies: Universities often reinforce and update their academic integrity policies to explicitly address the use of generative AI. This includes guidelines on plagiarism, proper cita etc…….
It is interesting that Claude prompts for additional information where as chatgpt acknowledges it cannot answer but also gives a generic HE response. Similar to a previous post, would I be compromising restricted access to these documents if I uploaded them for analyses?
I had to think carefully about what texts to use for this task. Are they in the public domain, do I have or need permission to use them in this way? I was glad to read Susan Murray’s post and her question to Claude about the storage of uploaded documents and I note the response. It was very useful and answered my concerns. Thank you Susan.
When I went to do the task I couldn’t set up a Claude account, it’s not available in my region – Ireland. I’ll have to try again soon!.
I uploaded a 4 page fact sheet which includes definition. (Admittedly, a challenging kind of document for this purpose but I wanted to test it.) Claude missed the definitions entirely, which are really key to the text. Claude’s summary was useful for summarising the history of the issue outlined on the document, but on the whole it missed the actual point of the fact sheet.
Professionally, I do wonder what the point of such a tool is? Aren’t we familiar enough with academic reading practices that we know how to find the information we need quickly? Isn’t there a risk Claude misses the nuances, as other commenters have pointed out? Perhaps summarising meeting transcripts is useful enough, but isn’t there also a risk that tone and the subtleties of points made in context etc will be missed by the AI?
In the classroom, Claude could be really useful for exploring the importance of reading and comprehension skills, as well as effective note taking practices.
Hi – hope not to late to this daily task!
asked Claude to summarise a document about creating a key practice in my field and
it shot out a clear and precise response super fast. it wasn’t an especially challenging
academic paper however it made the points it easier to understand?
yes AI misses the nuances in creating ART but for this task big thumbs up for me.
Not too late – glad to see you here!
Just catching up with Friday’s task. I’m new to using AI tools, so great to be pointed towards another one to try out. The first PDF I tried to upload to Claude was too long, but it produced a nice summary of the next one I tried. I’m learning so much from the discussion – amazing that you can generate an AI blog post from a Teams transcript!
Claude is a new one to me, liked playing around and on the whole reasonably accurate
I asked Claude to summarise a pdf of a scientific publication, and the answer was ‘Due to unexpected capacity constraints, Claude is unable to respond to your message. Please try again soon, or get notified when paid plans are available’
Is this because it is unable to deal with complex data?
Thanks
My experiences using Claude (summarised by me, not Claude :-P)
How did you get on with the task – did you find it useful?
I hadn’t heard about Claude before and when I discovered it through this course, I thought it could be very useful, but it could also be dangerous.
Its usefulness is evident through its ability to create cecent summaries of .pdf and .word documents. This can be very useful for people with dyslexia or other learning differences, who have issues with memory retention in relation to reading etc. and it can also help with note taking.
However, I am not sure how accurately Claude captures the information or if it even focuses on specific aspects of knowledge the document seeks to disseminate.
Overall, I do think Claude can be useful, but I wouldn’t feel confident in only reading the summary without intending to read the text myself later on. Maybe instead of a substitute it can be a complementary tool used before or after engaging with a document, to ensure that key points are taken away.
Summarising and paraphrasing text is a key writing skill that we encourage our students to develop – does this use of AI undermine this skill?
I wouldn’t say undermine, but it can certainly hinder someone’s abilitu to summarise and paraphrase, if this tool is used passively and exclusively to provide sumamries. Also, I don’t know how ethical it would be to copy and paste sumamrised text and include it assignments. I would personally use it as a supplamentary tool.
Should we be using it and promoting it with our students?
I would certainly use Claude with long documents of meeting notes, to help me keep track of actions needed and to ensure I’ve double-checked my understanding of things discussed. I don’t know that I would promote it to students, for the ethical reasons discussed above, but also I don’t know that I’m the appropriate person to make this call.
The summary was OK although I occasionally found the ‘notes’ style hard to read. I can’t imagine using it in day-to-day practice but might be useful for a ‘first draft’ for an abstract or a resource description
I asked Claude to summarise a well-known paper on interactive multimodal learning as a checklist of tips for creating online learning. It identified key points and created a checklist of 10 items. Then, I asked for an example of how to use a ‘pedagogical agent’ for guiding students and it gave a useful example. I liked the way it was able give a practical example of the concept.
I think it would be better to ask students to attempt their own summary of a text first, after being taught the academic skill, before using an AI as a second perspective or comparison. They would also need to be made aware of possible copyright implications of uploading other people’s work.
I used it to summarise a scientific journal article and it did a good job of it (it was one of my own, so I could verify it’s accuracy!). I think this could be really useful for students to get the gist of primary scientific research paper, which can be challenging at an early stage their your studies.
The ability to summarise Teams meetings, or similar, is brilliant for students (or anyone) who may struggle with this due to an additional support need, and arguably allows for multiple means of engagement (CAST guidelines).
Hi, some of my students have used this to better understand academic papers… My issue is with what happens we upload. It all goes into the system and we will not have the permission to do this… what are the ethical implications or should we not to worry about this?
Claude summarised a book chapter pretty accurately. It is something I can see my students using, so yes, it does undermine students’ capacity to understand and summarise a text.
The task was incredibly helpful. Ive never used Claude before so this was my first time! It not only saved me time but also provided accurate and concise summaries that captured the essence of the content uploaded (a PDF of a previous conference agenda i attended recently).
I think that AI can be a great tool to help encourage can offer a different perspective and help students refine their summarization abilities through comparison and analysis.
I think students should be able to use this when and where called for within the HE setting. I like to see the positive and feel that AI can enhance their efficiency and encourage a deeper understanding of the material/subject are. It serves as a modern and practical approach to learning.
As mentioned, this was my first time using Claude. I found Claude’s summaries to be quite good and i’m sure my opinion of this will change over time! (Hopefully, it’ll continue to be a positive experience!).
Blog post:
Claude Summarised by Claude:
https://totallyrewired.wordpress.com/2023/12/06/claude-summarised-by-claude/
Not sure why Claude needs you to verify your account by mobile number? Does anyone know? I would’ve thought email would be sufficient. This is a privacy concern for me.
For the task, I got Claude to summarise a wellbeing support document I’ve been referring to and it did it pretty well! I’m impressed. It’s my first time using Claude. It was a bit formal so I asked it to “make that summary less formal”, and it went a bit too much the other way. The result was better when I said “Make that more professional but friendly.” So it got there in the end! I might look into that Coursera prompt engineering course to help me specify everything from the beginning to save time.
I think some students could begin to rely on tools like this and there’s a danger they wouldn’t develop the skill of summarising text themselves, which could be problematic in exam and future work contexts. It probably needs to be carefully mentioned (and maybe AI responses could be critiqued and improved on by students as a task to get them to think critically about their work and AI output).
Getting Claude (or ChatGPT) to summarise meetings from notes/documents would be helpful, both to recap and for those who couldn’t attend.
Claude claims it’s to prevent spam and abuse. https://support.anthropic.com/en/collections/5845817-claude-ai-phone-verification
Checked but not available in Canada.
Sorry