Assessment methods at top 10 unis - movement away from traditional exams?

[Springflowers1]

Really interested in getting feedback and experiences on this for humanities/social sciences at top unis as DC2 is in yr12 and I think this is something we should be more aware of when evaluating options and comparing unis. For context DC1 is at Oxbridge doing an exam heavy essay subject with very little scope for coursework or other forms of assessment. First 2 years is all very stressful traditional exams and 3rd year is 3/4 traditional exams plus a dissertation or similar. Uni did do 5hr open book exams for a while after covid but reverted back to exam halls due to concerns about cutting and pasting of old essays and use of AI.

First time around I wasn't fully aware that so many other top unis now employed many diverse ways of assessing students and that often the majority of exams were open book eg 24hr ones?

Those of you with DC doing essay subjects at high ranking unis, how have they found these alternative modes of assessment, particularly 24hr open book exams? Does the removal of the need to memorise everything significantly reduce stress levels? It makes alot of sense to me - nothing in the world of work is about memorisation of vast amounts of information, but I know it's not straightforward because of issues with AI.

DD has only had exams in her maths heavy modules. Assessment for the humanities side have either been standard essays and final assessments was a 48hour open book exam where they had to do 2 2000 word essays.
I liked the open book idea on unseen questions they had to find all their references from scratch and put together a coherent argument. And juggle two very different questions. Im sure AI could have pulled something together, but I would think it would be obviously too polished.

I think it depends on what people mean by use of AI. I’ve actually just done a university course myself and one of the lecturers was positive about using AI as a research tool. In that respect I think it’s completely acceptable. Having done my undergraduate degree 30 years ago and my Masters 15 years ago I can see why standards have improved. Access to resources is incredible compared to the past. I could spend a day looking for books which barely helped me compared to now when within 5 minutes I have every resource I need. There is a world of difference between this and pressing a button asking AI to write an essay on a particular topic.
The professor I heard speaking about this was in two minds about students reluctance towards AI, not because she was happy with them cheating their way through a degree but because these students may get a job which requires some use and knowledge of AI in the future. For these young people it’s like asking vegans to eat a bacon sandwich. Interesting a relative of mine who studies in the US was told by one of her professors that AI won’t replace them but a student who knows how to use AI may do. I thought that was an interesting take on things which may explain the UK professors worries.

AI is a major problem and unis are so behind the curve. We have gone back to in person exams at present although we had been moving away. But dissertations are still a major headache. It's not that we can't see AI influence - problem is there is nothing we can do about it. And some use of AI should be acceptable - but how much? Dissertation supervisions mean that students are effectively pre-vivaed so we get some kind of bench mark but you can't fail someone for AI editing and you can't prove AI did the conceptual thinking. You can sometimes spot lousy source use via AI - but that is improving all the time.

Yes, in real jobs you use AI - but the baseline understanding needs to develop via personal engagement. You can't get that by outsourcing all the content to an external brain.

We have to absorb the use of AI somehow - but we are long way from doing that effectively at present.

The pp who said most students won't use it - that is absolutely not my experience at all.

Sadly, the risk of AI use is huge. Students use it to study, write etc. Summaries of articles, getting at the very least a starting point from AI on an essay rather than ploughing though reading through things that are hard and facing the blank page and having to stick it out until you come up with something. The values of efficiency and outcome rather than process and resilience are prized no doubt in part due to the high cost of education, CoL crisis and fears about the future.
Lots of unis are going back to exams that have a live element eg timed written exams, oral presentations etc to protect the integrity of the degree as otherwise they will start to become pretty meaningless
Open book 24 hour exams can potentially be helpful re not having to memorise every minute point but it is a false security to some extent because you still need to know the material really really well. Students often also interpret a 24 hour exam as meaning they should work for literally 24 hours no matter how much they are told it is not supposed a physical test of endurance. This has a knock on effect on other exams where lack of sleep means anxiety rockets.

I think you are referring to me. I wouldn’t say most student won’t use it. I do believe most students wouldn’t use it to cheat and have AI write their actual essays. I have used it to summarise texts I need to write my essays. I haven’t used it to actually write them. This is an important distinction.

@ProseccoPiewhat are you talking about? Do you do exams as part of your work? Probably not. Exam pressure already exits through GCSEs and A levels so the students, if at top unis, will have already proven they can deal with that kind of pressure. And any student would certainly not rely on Google for answers. You must site all your sources and Google ain’t coming to be one of them!
My DD is studying Sociology at Durham. The majority of her assessments are from essays, with just one exam this term (Y2). This is far more sensible as it mirrors what will be required should she continue in this field. They are taught theory and statistics and while it’s a fact based subject - I’m not sure how an exam in the traditional sense would be useful. Sociology requires the researcher to answer the why, requiring thought, analysis and looking at deeper nuances. The student must put forth a logical argument rather than having a ‘correct’ answer. If a student uses AI that’s to their detriment. Why bother going to uni at all if not to discover and write in their own words?
I also believe A levels should be scrapped. I think continual assessment with a wider curriculum is a far fairer and more sensible way to grade students (and termly tests can be standardised to make it comparable across schools). This system works in other countries and leaves more options open not to mention offers would be unconditional!

I agree that A levels should be scrapped. It values the wrong type of intelligence and gives an advantage to private school pupils.

What's the 'wrong kind' of intelligence @Rbof ?

The human mind requires a body of content to develop problem solving skills - it cannot develop those skills if the content is on an external drive. It is, however, very understandable that students under pressure for whatever reason might seek help to varying degrees. I think a lot of people don't understand what is possible with AI these days - there's a long spectrum from summarising a few texts to actually writing the essay. Gemini is very good at conceptual analysis and evaluation for example (not that I want give people ideas...). Students do use AI - a lot. I've had quite a bit of AI training myself and I'm thinking about how to approach this with students. How can we teach them to use it OTOH and not compromise their education OTOH.

It is perhaps also worth adding that students at Oxford and Cambridge have the privilege and benefit of weekly tutorials / supervisions. They therefore also get lots of experience doing research and writing essays / other work problems under pressure that better mimics the real world in fast paced workplaces (3 x 24 = 72 essays over three years - often more than 1 tutorial per week). I appreciate these don't count for the degree but they do develop excellent analytical and writing skills and a huge opportunity to learn to receive feedback.

I went through 'the only your last 10 days of finals' counts at Oxford 30 plus years ago. I'm not convinced it is the best way to assess students even though I did quite well. DC1 two years ago had 6 exams, a dissertation and two extended essays so finals counted for ~80% of his degree. DC2 next year has seven final exams and a dissertation so final 10 days counts for about 85% of their degree.

Whatever system you have will suit some learning styles better than others but a total focus on one set of exams in your final year is in my view brutal and unfair. Dreadful if you are unwell or a close family member falls ill / dies.

@Juja totally agree. My DC1 is 100% final year with a max 25% of that from a diss or extended essays. 75% comes from 3 exams. It’s horrendous pressure. It is also alarming that there is a significant disparity between men and women in terms of % 1sts because of this.

Someone who is good at regurgitating information but struggles to develop their own ideas.

@Rbof
But have a good memory is important when you are a doctor dealing with an emergency situation, a lawyer arguing a case, a teacher being asked a question in a class, a computer scientist knowing how different systems interact, an engineer doing on site problem solving, a consultant being asked for advice on a complex matter, an accountant being asked by their client about regulatory rules or taxes. In these situations, these professionals can't just go and ask AI for the answer.
I think exams are an important part of university assessment particularly in STEM.
My lawyer friend has said that AI can hallucinate cases so if a newbie doesn't know the cases well enough, they can't spot it.

Yes a good memory is useful but very much only part of what constitutes intelligence. The examples you give all need some ability to think independently and creatively.

While obviously the tutorial system develops skills that students at universities other than Oxbridge are far less likely to develop, there Is an unfolding issue with essays at Oxbridge that have to be written in a short time frame drawing on canonical sources. Some UK humanities courses now require students to write long research essays that draw from multiple sources and are closer to a Harvard essay than the traditional Oxbridge essay in scope and depth. Such research essays are also AI proof to an extent that an Oxbridge essay, where the quality of the argument rather than the scope of the research is the KSF.

I have also read of academic research that points to longer summative written work being better for intellectual development than shorter formative pieces.

Yes, which is why PhDs are tested on very long pieces of writing.’

Exams aren’t the only way to learn though

No university worth the title is allowing assessment entirely by written coursework now. International students have long regularly paid someone to do it for them and AI has democratised the same effect for all.

That’s a ridiculous statement to make. It assumes every university student is a cheat, which is simply untrue.

You definitely don't work in a uni then. Any student who doesn't use ai now is at a massive disadvantage. You can't assess fairly if AI is an option.

The aggressive naivety of some people is literally infuriating.

There is nothing naive or aggressive about my statement. I actually tutor in a university. I have recently finished a university course. My DP and 2 DC are currently students at top 2/top 10 universities. I am speaking from experience. I understand that students use AI. That doesn’t necessarily mean they cheat. It’s a useful tool used in the right way. Pressing a button to ask it to compose an entire essay is quite a different matter.

It means you can't tell who is cheating and who isn't and anyone who doesn't use ai is choosing to come second. Get a grip!

You must be one of the external jobbing tutors to have such a view of the current position

Anyone paying English tuition fees to send their kids to a cheat's paradise is a pliant mug

The challenge comes if using assessment styles developed pre-AI in the post-AI world without additional guard rails.