I like to think of myself as generally optimistic. Like everyone, I make jokes about the machines taking over but, by and large, I feel excited by the potentials of AI (whilst keeping one hand behind my back tightly gripping a broken bottle just in case). In higher education the most pressing fears relate to threats to academic integrity and, as much as some might like, this is an issue that can’t be addressed with a broken bottle. Instead, it’s forcing/ encouraging/ catalysing (choose as per your perspective) us to reflect on how we assess students (see AI+AI=AI) . Whilst ChatGPT caught the world’s attention and millions of educators and students went slack-jawed and wide-eyed at first exposure (most soon of course started to see how frequently it ‘may occasionally generate incorrect information’ such as in this experiment I did). Whilst ChatGPT allowed us to draw our own (often erroneous) conclusions about its capabilities, other tools built on the same parameters as ChatGPT are, in different ways, honing in on the ‘help with essays’ market. On the one hand, this may be the death of the essay mill (who, after all, once they have decided to go down an academic dishonesty path, would decide to pay an essay mill to do what anyone with an internet connection can do for themselves?). On the other, it really blurs the picture of what is honest or dishonest or what we at UCL frame as ‘unfair advantage’
Two tools have caught my eye recently and in this post I want to reflect on the different impressions they give in terms utility and ethical use for students wanting to write an essay.
Chatsonic is one aspect of the ‘writesonic’ suite of tools – reviewed in considerable detail here with pricing and potentials. It caught my eye because of the bold claim to have addressed ‘limitations of chatGPT’ and to be ‘the best alternative’. I thought I’d ask it why it was better and I feel like the responses do a lot of talking for themselves:
Prompt: why is chatsonic better than chatgpt?
ChatSonic is a powerful AI writing assistant that offers superior accuracy and speed compared to ChatGPT. Unlike ChatGPT, ChatSonic produces 100% original content that is thoroughly checked for plagiarism. In addition, ChatSonic offers detailed analytics and feedback to help you track your progress in writing and make improvements. ChatSonic also provides an extensive library of templates to save you time and energy when you’re writing, allowing you to focus on the quality of your content. Finally, ChatSonic is backed by a team of expert writers who can help you refine and polish your work to perfection.
Prompt: Explain the process you use for plagiarism checking
ChatSonic uses a combination of automated checks and manual reviews to ensure that all content produced is free of plagiarism. Our automated system checks the content against the web and other sources to detect any potential plagiarism. If anything suspicious is detected, it is flagged and sent to our team of expert writers for a manual review. We also use specialized tools such as TurnItIn and Copyscape to detect any possible plagiarism. All content is thoroughly checked to ensure it is completely original before it is delivered to our customers.
Prompt: If I use chatsonic to write an essay will I get into trouble at my university?
No, you should not get into trouble when using ChatSonic to write an essay. Our content is completely original and plagiarism-free so it should not be flagged as such by your university. However, it is always a good idea to review the essay yourself before submission to ensure it meets all the requirements of your assignment. Additionally, make sure the content you create with ChatSonic is unique and tailored to your own voice and style, otherwise it may be noticeable that you have used a writing assistant.
To say this makes me uncomfortable is a real understatement. It only tangentially addresses the wider integrity issues and in its first response is pushing me to conflate plagiarism (and its avoidance) with those wider integrity issues. Two posts in its resources section are worth noting. This one on plagiarism reinforces the message the chatbot produced. This post on replacing writers asserts that AI is an assistant but I couldn’t find much more detail on the ethical dimensions.
Maybe though I was ‘unfair’ (is that the right word?) with my prompts so it’s worth looking at some of the other functionality of the wider writesonic suite. The writesonic suite offers a dazzling array of tools for a range of purposes from website copy to social media content:
I was keen to look at the ‘citations generator’ as this is an aspect of ChatGPT that is a recognised weakness. You can use a URL prompt and a text based prompt. The text based prompt I used was itself generated in chatsonic. It takes the text in the linked article or whatever you paste in and identifies ‘facts’ with suggested citations. The web articles are mostly relevant though the first journal article it suggested was a little off the mark and I’d need to be lazy, in a massive hurry or ignorant of better ways of sourcing appropriate resources to rely on this. At this stage!
The second tool that I noticed (via the prolific AI researcher Mushtaq Bilal) was Jenni. The interface is well worth a look as I feel as if this foreshadows what we are likely to expect from generative text AI integration into tools like Microsoft Word.
The first thing I noticed, however, is the blog with the most prominent word across posts being ‘essays’. Each is designed to address an approach to a different type of essay such as the compare and contrast essay. It offers clear suggestions for different approaches, a worked example and then, right at the end says:
“ If you want your compare-and-contrast essays done 10x faster, we recommend usingJenni.ai along with the tips and guidelines we provided you in this post.Jenni.ai is a fantastic AI software that aids your essay writing process so that you could produce your writing needs faster and better than ever.”
Another post deals head on with the ethical and integrity issues of using AI to help write essays and makes a case for use of ‘AI as a tool, not a solution’ where the goal is a “symbiotic relationship between the critical thought of a writer and the processing speed of AI”
The tool itself, unlike the huge range of offerings in Writesonic is a relatively uncluttered interface where you start by typing a title, it offers a judgement and suggestions if appropriate.
In addition, it offers in-line suggestions from whatever has come before. The prompt engineering continues through what feels like a single document rather than a chat. If you don’t like the suggestion you can get another. Here I typed a prompt and these are the first three options it gave me. Note the positive aspect on my worried prompt in every case!
My prompt sentence….The worry is that these tools will be used by students to cheat
but the reality is that most AI tools are designed to help students learn and improve their writing skills. [first suggested Jenni response]
The worry is that these tools will be used by students to cheat
on their assignments by generating essays automatically, but they can also be used to provide feedback on areas where students need…[the second option]
The worry is that these tools will be used by students to cheat
but their true purpose is to assist in the learning process by providing immediate feedback and identifying areas where improvement is…[third option]
The other noticeable aspect is the option to ‘cite’ – Here it offers a choice of MLA or APA 7th and the sources are, unlike ChatGPT’s famous hallucinations, genuine articles (at least in my limited testing). You can select ‘websites’ or ‘journals’ though I found the websites tended to be much more directly relevant than the journals.
I really have only just started to play with these though and new things are popping up all over the place every day. Most educators will not have the time to do so though. Students may see and use these tools as an extension of those they use already for translation or improving writing. The blurry zone between acceptable and unacceptable is getting more ill-defined by the day.
What can I conclude from this? Well, firstly, whatever the motivation on the continuum ranging from ‘give us all your money’ to ‘I believe the children are our future’, the underlying technology is being adapted rapidly to address perceived limitations in the tool that has brought generative text AI tools to our attention. We may not like the motivations or the ethics but we’ll not get far by ‘making like an ostrich’. Secondly, It’s not good enough for us (educators) to dismiss things because the tool that many are now familiar with, ChatGPT, makes up citations. That’s being addressed as I type. The number of these tools proliferating will soon be too huge to keep a decent handle on so we need to understand broadly how discrete tools might be used (ethically and unethically) and how many will integrate into tools we use daily already. In so doing we need to work out what that means for our students, their studies, their assessment and the careers our education is ostensibly preparing them for. Thirdly, we need to open up the discussions and debates around academic integrity and move on from ‘plagiarism’ as public Enemy No 1. Finally, where there are necessitated changes so there are resource implications. We need to accept that to prepare ourselves, our colleagues and our students we will need to adapt much faster than we are used to and properly resource however we attempt to address the challenges and opportunities ahead.
Note: This post is not an endorsement or recommendation of any products mentioned and should be read with that clearly in mind!
If students use AI to produce an essay, would that count as their own work? I’d say No.
So then, what would they be marked on? How do we get around this easy way out or “cheating” issue?
in my view, we have to really work out whether we all agree on what constitutes cheating first! Then on where we draw the line on ‘assistance’ this may well be something that has to happen locally in departments (or even on specific programmes). In my view, ‘use AI to…’ could mean a lot of things. We currently use (integrated into MS offcie for example )forms of AI to assist so the difference is the extent. The key surely is where is the evidence of learning?