All 4 entries tagged Feedback

AI Feedback Systems: A Student Perspective – Mara Bortnowschi

The buzz is endless – AI is taking Higher Education by storm. Since the launch of ChatGPT, everyone seems to have an opinion, and rightfully so. It’s so new and we have yet to fully understand its potential and the impact it will have. Within academia, the general sentiment mostly surrounds concern for responsible use, as many students have heard their professors and lecturers warning them against its use. However, its rapid growth and widespread adoption demonstrate that it’s not going anywhere soon so instead of avoiding it, it should be given the time of day to understand the risks and challenges but also the opportunities it presents. Furthermore, I think the student voice in these discussions has been underrepresented, but really students can be the key to harnessing this technological advancement as an asset to enhancing learning and education.

The WIHEA group have already explored a number of subjects regarding AI in education from student perspectives that can be found on the group’s Artificial Intelligence in Education webpage. These have included emerging questions AI presents, the risks and ethics of academic integrity, evolving assessment styles to mitigate and integrate AI into assessment and how teaching may evolve. I will now explore some of the opportunities that are being presented with the widening availability and access to AI tools for students to enhance their learning and to generate formative feedback. While summative marking has been and continues to be required to be marked by human markers according to the UK Quality Code in Assessment (UKSCQA, 2018), formative feedback has more flexibility, and we are now presented with an opportunity to test and utilise the capabilities of these AI technologies in providing timely, constructive, and developmental feedback.

Existing feedback systems

This notion will be particularly explored with regards to formative elements of summative assessments. Feedback should allow a student to understand strengths and weaknesses of their work and if engaged with effectively, can be used to improve academic performance, and thus learning. Especially throughout the pandemic, we have seen the role of feedback change massively: as more of education has shifted online, reliance on formative assessments has increased as assessments for learning. This is in contrast to summative assessments which more so represent assessments of learning (Wyatt-Smith, Klenowski and Colbert, 2014). Formative assessments also are an opportunity for autonomous learning by developing one’s own skills and relying on self-motivation. It would also be fair to say that formative feedback can be self-assessment of sorts, as even though the formative feedback is generated externally, it is the engagement with, and learning you apply from it that will ultimately make a difference in each student’s performance.

AI generated feedback

So what could incorporation of AI in these feedback systems change? Well, the use of algorithms in generation of feedback is not an entirely new concept. Algorithms, such as Grammarly and Sketch Engine, have been around for a while and they can generate feedback on academic writing and are for the most part freely available, or students are granted access to them by their institutions. But with more complicated algorithms that use machine learning, we can apply them to provide specific and personalised feedback. To make this even more applicable, by integrating what could be different elements of summative marking criteria or rubrics, they could provide some of the most relevant feedback at a moment’s notice.

This application is indeed being explored right here at the University of Warwick. Isabel Fischer, a WBS professor, is trying to pilot a deep learning formative feedback tool that has the potential to provide more comprehensive feedback that was developed with WBS marking criteria at its core. By simply submitting a pdf or word document, the algorithm instantly produces a document of in depth feedback on the four aspects of WBS marking criteria. This could be just the start of developing similar department-specific feedback tools taking into account department-specific assignments, marking criteria, and writing styles for drafts of academic writing. While there are definitely some considerations to look out for, this is fascinating and shows great promise as a tool to increase student autonomy in adapting how they might approach assignments to still have the opportunity to personally benefit from formative feedback.

Considerations of using generative AI

The considerations I mentioned earlier are worth discussing as students are turning to generative AI technologies like ChatGPT more and more. While these technologies are being developed to simulate human intelligence, there are some things they are simply not capable of. For example, it lacks expressions or expressive language. If using them to generate feedback on your writing, you should be aware that they will not always be able to grasp the nuances or expressive language in that writing. In other words, any feedback you receive from AI should be approached critically. You decide what you implement from feedback you receive, and you are responsible for identifying and understanding what truly can improve your work. This is all part of the responsible use of AI, but really also goes for human generated feedback. Your assignment at the end of the day will still be marked by a human marker with in-depth subject-specific knowledge and skills that they are asking you to learn and demonstrate in your assignment. I think this is the quick, irresponsible and neglectful way people have been quick to exploit resources like ChatGPT, where they do not doubt any response it has generated and implement them into a piece of work, only to find that its references are wrong or entirely don’t exist. Firstly, this should not be the way we utilise it, as this is blatant plagiarism, but secondly, a critical approach should be used to (for example) verify references, and critically understand that the way AI answers can lack certain elements of context. Regardless, the point still stands: responsible applications of AI technologies should not be about using it to do your work, but using them to enhance or improve your outputs.

Engagement with AI technologies and feedback

A new level of engagement with AI has been seen since the release of ChatGPT and DALL-E. Perhaps this is rooted in the great advancement that this represented or, more sinisterly, the opportunity to exploit the technology to minimise workload. Regardless, everyone’s interest has been piqued, and the level of engagement has been massive, exceeding what anyone might have expected particularly from students. At the University of Warwick alone, students have made a total of 850,000 total site visits to ChatGPT in the first two months only on the university’s Wi-Fi (SWNS, 2023). I think it’s important to try to understand why this might be in order to be able to channel this traffic for good rather than just fear this alleged ‘cheating epidemic’ that the media has been dubbing it.

In contrast to the older algorithm technologies that have been around, like for example the previously mentioned Grammarly and Sketch, which experienced much more moderate levels of engagement and use. Reasons vary from lack of awareness, to limited breadth of feedback to language, or to lack of confidence in the feedback they provide. AI has surpassed some of these limiting factors in that it is capable of generating a wider breadth of feedback that can include language, style, structure and more. The confidence in the feedback it produces is reassured by the continuous concern from educators. If professors are afraid AI technologies can be used to write entire assessments, then they must be capable of doing so.

Further benefits

As a result, we have seen students be a lot more open to trying to use ChatGPT, and I think we should utilise this eagerness in a way that encourages students to enhance their academic achievements. By introducing resources such as Isabel Fischer’s feedback tool or teaching students how to formulate prompts for ChatGPT to generate similar constructive feedback, we can guide a smooth integration of ChatGPT into Higher Education practices. And there are so many more benefits we have the potential to see. For one, this alleviates a massive workload off staff. If such tools are available to take care of the writing style and structure, staff’s role in formative feedback can remain more focused on content. The speed (or even instantaneity) with which AI can produce feedback also makes feedback more accessible. Furthermore, students can engage with it as many times as they like, inputting multiple drafts, as they are less limited by staff work capacity. Also, different students work on different timescales and with different approaches when faced with an assignment. This further widens accessibility to students that might start assignments later that what might normally be a formative deadline. Communicating these advantages is key in order to achieve these outcomes and to harness this technology towards enhancing educational experience for both staff and students.

Conclusion and personal experience

In my experience thus far with using ChatGPT, I have had mixed feelings. On the one hand, I am very apprehensive of the fact that its use is quite contentious at the moment, with some professors explicitly prohibiting its use or consultation. On the other hand, it is a resource that is available and it feels foolish not to use what is at your disposal. Throughout the research that went into this article and discussion with faculty members about its potential to provide feedback, I have been introduced to a very clearly constructive way to engage with ChatGPT, that seems to make both staff and students happy. While we are still in the early stages of understanding the potential and risks of generative AI technology, at the end of the day this is a tool that will have implications for Higher Education and we are being faced with the possibility of either embracing it, in various ways such as to generate formative feedback, or let it escape our control at the cost of academic integrity, because it is clear that prohibiting its use will not prevent people from exploiting it.

For further queries: marabortnowschi@yahoo.ca or mara.bortnowschi@warwick.ac.uk (may expire soon)

Reference List

SWNS (2023). University of Warwick fears cheating epidemic as data reveals huge number of students visiting AI website during exams. [online] Kenilworth Nub News. Available at: https://kenilworth.nub.news/news/local-news/university-of-warwick-fears-cheating-epidemic-as-data-reveals-huge-number-of-students-visiting-ai-website-during-exams-176836 [Accessed 19 Jun. 2023].

UKSCQA (2018). UK Quality Code for Higher Education Advice and Guidance Assessment. [online] Available at: https://www.qaa.ac.uk/docs/qaa/quality-code/advice-and-guidance-assessment.pdf?sfvrsn=ca29c181_4 [Accessed 16 Jun. 2023].

Wyatt-Smith, C., Klenowski, V. and Colbert, P. (2014). Assessment Understood as Enabling. The Enabling Power of Assessment, pp.1–20. doi:https://doi.org/10.1007/978-94-007-5902-2_1.

Abigail Ball : 10 Jul 2023 09:00 |  Tags: AI Assessment Feedback Mara Bortnowschi |  Comments (0) |  Close comments |  Report a problem

Using AI for Formative Feedback: Current Challenges, Reflections, and Future Investigation

By Matthew Voice, Applied Linguistics at the University of Warwick

One strand of the WIHEA’s working group for AI in education has focused on the role of AI in formative feedback. As part of this strand, I have been experimenting with feeding my own writing to a range of generative AI (ChatGPT, Google Bard, and Microsoft Bing), to learn more about the sorts of feedback they provide.

The accompanying presentation documents my observations during this process. Some issues, such as the propensity of AI to ‘hallucinate’ sources, are well-documented concerns with current models. As discourse on student use of AI begins to make its way into the classroom, these challenges might provide a basis for critical discussion around the accuracy and quality of the feedback produced by language models, and the need for student to review any outputs produced by LLMs.

Other common issues present different challenges for students using LLMs to elicit formative feedback. For instance, the prompt protocol in the presentation revealed a tendency for AI to provide contradictory advice when its suggestions are queried, leading to a confusing stance on whether or not an issue raised actually constitutes a point for improvement within the source text. When tasked with rewriting prompt material for improvement, LLMs consistently misconstrued (and therefore left absent) some of the nuances of my original review, in a fashion which changed key elements of the original argumentation without acknowledgement. The potential challenges for student users which arise from these tendencies is discussed in more detail in the presentation’s notes.

In addition to giving some indication of the potential role of LLMs in formative feedback, this task has also prompted me to reflect on the way I approach and understand generative AI as an educator. Going forward, I want to suggest two points of reflection for future tasks used to generate and model LLM output in pedagogical contexts. Firstly: is the task a reasonable one? Using LLMs ethically requires using my own writing as a basis for prompt material, but my choice to use published work means that the text in question had already been re-drafted and edited to a publishable standard. What improvements were the LLMs supposed to find, at this point? In future, I would be interested to try eliciting LLM feedback on work in progress as a point of comparison.

Secondly, is the task realistic, i.e. does it accurately reflect the way students use and engage with AI independently? The review in my presentation, for example, presupposes that the process of prompting an LLM for improvements to pre-written text is comparable to student use of these programmes. But how accurate is this assumption? In the Department of Applied Linguistics, our in-progress Univoice project sees student researchers interviewing their peers about their academic process. Data from this project might provide clearer insight into the ways students employ AI in their learning and writing, providing a stronger basis for future critical investigation of the strengths and limitations in AI’s capacity as a tool for feedback.

This is blog 14 in our diverse assessment series, the two most recent previous blogs can be found here:

• Assessments: Capturing Lived Experience and Shaping the Future
• Building knowledge on the pedagogy of using generative AI in the classroom and in assessments

Abigail Ball : 26 Jun 2023 09:00 |  Tags: Assessment Feedback Formative Matthew Voice |  Comments (0) |  Close comments |  Report a problem

Trainee teacher 7: Varying degrees of success – Robert

Although I can sometimes grumble and moan about the extra workload of reading research papers, there can be no doubt that I’ve taken ideas and inspiration from the research I’ve written. Over the year so far, I’ve been trying to take ideas from my reading and implement them in classrooms, with varying degrees of success.

From an early stage, my research into theories of effective marking and feedback yielded some ideas that I could implement into my classroom. Reading work by Hattie and Timperley gave me the idea to use displayed success criteria for written tasks in class. This allows students to constantly self-assess their work as they go, and continuously generate next steps at any stage during the lesson. This allows me to use the criteria to also structure my written feedback, saving a lot of time thinking about what next steps are necessary. I recently took this to another level in terms of student engagement with the criteria, by challenging students to mark a sample piece of work before they started working on their own task. I was impressed with the levels of critical engagement with the criteria, and by their willingness to judge whether a piece of work was worthy of meeting a criteria rather than just being present. The feedback was very focused, and they were very demanding of detail (in hindsight, maybe telling them I had produced the work was a bad idea!)

More recently, I have been reading into theories of discovery learning and how to make it effective in the classroom. The overall picture I obtained from my literature review was that discovery learning can yield improved outcomes, but not in all cases. I identified two main factors that can help discovery learning be more effective:

Choose the class carefully. When results were separated out by student ability, a positive effect on higher ability students was observed, often masked in mixed studies by a negative effect on lower ability students. Using this strategy appropriately is an important facet of making it successful. In terms of implementing this, I have primarily used these activities with only one of my classes, where all students are targeted an A or A* grade.

Don’t just leave them on their own! Some studies gave students as young as 7 no guidance, and expected them to be able to learn. There is no way that this would yield effective progress and learning at almost any age As a result, when I have used discovery-style activities, I have always provided scaffolding questions and walked around providing support to ensure that students have the supported environment to allow them to make those discoveries.

The opportunity for discovery learning to produce improved outcomes has been particularly of interest to me, and I have been trying to implement more and more in my lessons where appropriate. Recently, I put on an activity where my students used dice to model radioactive decay. Using the structured worksheet, students were able to work through and calculate a half-life for their radioactive ‘sample’. This then led into a discussion of half-life, with students moving on to look at how it relates to carbon dating in their next lesson.

Abigail Ball : 04 Jun 2018 09:59 |  Tags: Effective Feedback Marking |  Comments (0) |  Close comments |  Report a problem

Marking – does research help? – Sean

Literature Review: How did reading around pedagogy affect my teaching practice? A trainee’s personal reflection

Certainly in terms of my own professional development, reading around effective feedback within marking revealed a lot to me which I had never thought of before, particularly the negative impact that feedback can have. Being a brand new trainee, I assumed that the more feedback the better for the pupils, so had no problem writing swathes of response for each piece of assessed work and spending a lot of my time in the process.

Initially when I first set foot into the world of marking, I wanted to attempt it on my own to see how I would independently respond to a piece of work; the result of which is aforementioned and this of course was unsustainable. I was recommended by colleagues to give lots of praise within my first set of marking as a way to build rapport with the pupils - and it definitely seemed to work, pupils appeared up-beat and engaged in the lesson which followed. Utilising this ‘praise culture’ fitted in well with the school marking policy of ‘two stars and a wish’, a principle used across many schools under various aliases; praise followed by ways to improve.

Being a relatively young teacher who was mistaken as a new year 12 student by year 13s does have its draw backs – you just don’t have the automatic respect which a mature teacher can assume from a class of students. In which case you need to adapt and use your strengths to build respect – this is where praise becomes invaluable and two stars and a wish offered me the opportunity to utilise this tactic. I enjoyed using this method as it gave me scope to praise the students and build rapport whilst also giving me the chance to comment on where they can improve. I assumed this was working well for me without giving it a second thought – I was ticking all of my boxes; praise, improvements and progress.

Praise is an essential tool within a teacher’s arsenal; however what became apparent within reading around my topic was that praise within feedback can have a detrimental impact on a student’s progress; studies have found that students can start to slack and relax when given praise on their work, removing their desire to push themselves further. Discovering this research has really changed my approach to the way I teach and particularly mark, however breaking away from giving lots of praise was something I struggled to do. Worrying I would offend some students about their work was a main concern; my thinking was that giving no written praise could in turn knock their confidence and impact their learning within future lessons.

Working on my new understanding of feedback, I have attempted various techniques to change my marking style; I still feel it is important to feature written praise, however I now use it in far smaller doses. Ultimately what I have taken away from reading around the pedagogy is that whatever principle you are researching, whether it be providing feedback or behaviour management; it should not dictate exactly how you teach but should instead add depth to your style. Use reading to mould your personal approach in a way which best suits you and your personality; the profession is based on all teachers having their own individuality and that is always important to bear in mind in your training year.

Abigail Ball : 12 Mar 2018 10:55 |  Tags: Feedback Praise-Culture |  Comments (0) |  Close comments |  Report a problem