How to Do a Literature Review with AI (Standard, Deep Review & Research Agents Explained)

There's going to be different times where you're doing a literature review with slightly different purpose. If you

are maybe in a research institute or somewhere that does data driven stuff, you might be doing something that that effectively is a literature review but maybe not necessarily as strictly called

that. And so I've got this list here and

that. And so I've got this list here and we can kind of look at this and go this this is roughly sorted by the depth and the rigor and the size and the effort.

Um, so I mean I guess those of you doing a PhD, the the good news and bad news is that your your literature your PhD is probably going to be the the largest,

longest, hardest, most in-depth uh literature review you will probably ever do. Uh, which part of the purpose of

do. Uh, which part of the purpose of that is to make you the expert in the thing you are getting the PhD in. Um,

but there's a whole range here. If we

step back a step from a PhD for an academic journal article, I need good rigor. I need to be very careful. Make

rigor. I need to be very careful. Make

sure that I've looked at the appropriate references for the thing I'm writing about, but the thing that I'm writing about isn't a 200page PhD's dissertation. It might be a a fourpage,

dissertation. It might be a a fourpage, fivepage journal article. So, I'm going to have to be very specific and very to the point. Uh I'm going to need to

the point. Uh I'm going to need to select things that are are much narrower. But then if we come a few

narrower. But then if we come a few steps back, if we think about things like policy work, grant writing, conference presentations, they still need really all of these. We we never

want to be doing a literature review and putting wrong stuff out there that's going to be really bad for you and for your career. Uh but we want to that

your career. Uh but we want to that level of depth and that level of effort is is going to vary depending on which one of these things there are. So very

I've kind of worked from the bottom upwards, but at the very top of the list there might be sometimes and in in my work sometimes there's there's times where someone might have maybe something about a particular medical procedure or

a particular disease or a particular kind of training component and go hey what's can we can we check a fact or can we check a claim or can we just get some

information about this thing and that still needs to be correct but I don't need to go to the huge amount of depth and the rewriting and the synthesis and all of that really deep

work that I would need to do for an article or a PhD. And so I present the present these here because we can think about size space and we can think about

how AI can help us in a range of different ways. So depending on which

different ways. So depending on which one of these things I'm doing is going to depend very much on which components of SciPace or or whatever AI tool I'm

using are going to be most useful and the way that I go about them. So I've

taken all those and then I've kind of just zoomed into three. So if I just have some sort of quick reporting task

then for all of these you can see that AI plays a role and I play a role. Uh

and at the bottom there I've said we should be thinking about this you should be using it to refine your thinking but you should never be using AI to replace your thinking. If you are just using AI

your thinking. If you are just using AI and copying and pasting stuff then you add no I can write a bot that does that.

I don't I don't need you to do copy and pasting. you need to be adding your

pasting. you need to be adding your brain and your value to whatever you're doing. So for all of these AI has a

doing. So for all of these AI has a role, you have a role.

Uh so if I just have something where it's some quick reporting, some some kind of finding and and just reporting upwards in the organization, something like that, I can jump into my AI tool. I

can get it to do the search. I'll say

here's my research question or the thing I want to find out.

It can do the search for me. It can do the summarization for me. I still need to verify. I'm going to look bad and I'm

to verify. I'm going to look bad and I'm going to risk my career if I just take that, pass it on, don't carefully check it. But I'll check it. I might edit it a

it. But I'll check it. I might edit it a little bit. I will absolutely

little bit. I will absolutely acknowledge that I used AI, but then I'm kind of done that that. So

for some of the things that I do now in my my research role where it's not the rigor of a journal article, it needs to be right, but it doesn't need that rigor. then sci-pace has been absolutely

rigor. then sci-pace has been absolutely amazing for stuff like that. For a

journal article, sci-pace is still going to be really helpful. I can do it's going to be more of an iterative search.

I'm going to search here's some good stuff. Maybe this leads to the next

stuff. Maybe this leads to the next thing. Do another search. Maybe I kind

thing. Do another search. Maybe I kind of adjust my research question. It can

help me a little bit with summarization.

It can also help me chat with papers so I can kind of understand the papers more.

um I we need to rewrite more deeply then with a PhD for the PhD we need to think more these AI tools they're going to help you explore ideas they're going

to give you structured support they're going to be really helpful chatting with the papers but there's there's intellectual ownership that absolutely has to be you and only you um and so you shouldn't be thinking this will

definitely be tools that will help it'll help you a whole lot I wish I had them when I was doing a PhD it would have saved many many many months of library

trips and misery. Um, but you are still needing to add quite a lot.

Um, and so I zoom in a little bit more on that here.

So those first two bullets, the reporting presentation, quick turnaround, I kind of described that.

Uh, journal article kind of described that. PhD, I've just got a little bit

that. PhD, I've just got a little bit more detail there. So, PhD, I'm going to be doing multiple searches. Your PhD is going to be several years long. So,

you're going to have to be doing searches and coming back later on to do more searches, see if there's new things that have been published. You might have slightly different changes of perspectives that you need to adjust

for. Um, SciPace has a find topics

for. Um, SciPace has a find topics function which is really helpful for brainstorming and narrowing down topics, particularly when you first get going.

Um, you will be getting information from other sources. So things like Zero where

other sources. So things like Zero where you've read some books, you've put that information into Zetero, um being able to combine that is going to be really

handy, but it's not you're not going to have all all of your um literature found in one spot for most of you. Um as

I mentioned before, the chat of paper is really helpful. Um, but I would be

really helpful. Um, but I would be thinking about the summarization and writing stuff here is it'll get you going. It'll give you some stuff to

going. It'll give you some stuff to think about. Um, but certainly for me,

think about. Um, but certainly for me, I've examined a number of PhDs. And if I received your PhD and it didn't look

like you wrote the literature review and it was just a bit of a a copy and paste uh from a summarization, then uh that that would not bode well for you. So

really important again and and I'm I'm sure there'll be probably 10 more times where I mention it is your brain and the human element really really important to stack on top of these tools. Um but

these tools are going to save you lots and lots of time.

Okay. So we are going to change what we're looking at and we are going to jump over to Sispace. So, scisspace.com.

Um, if you want to just give me a a nod or a indication in the chat that everyone can now see SciPace.

Yes. Excellent.

Cool.

Okay. So, what we're going to do is I'm just given a lot of you have never seen SciPace before or not SciCA users.

little bit of a kind of a general summary of stuff that's here and then uh I will show you the literature review

tool and then I'll show you the agents that will do literature reviews but they'll do more on top of the literature reviews as well. So here's our homepage.

Uh, and if you've used chat GPT or any AI, um, you you'll see that familiarity of here's the textbook text box in the start and we've got some little adjusting things around it and then some

more info. We'll come back and we'll

more info. We'll come back and we'll have a look at this. So, the homepage is where we get agents to do stuff. But

down the left hand side here, we can see there's actually a whole lot of different headings of different spots and different tools. Um, so my library, we can upload our own PDFs into the

system to get it to include an analysis and to summarize. Uh, my notebooks, so that's where we would save summarization in any of our own writing. Uh, the agent gallery is pretty cool. If we click on

that starts off by showing the featured, and you if you looked at these featured ones, you might go RNA, neuron, it's all it's all biology stuff. Um, but I think that's just cuz those are the most

recently added ones. So, they're kind of in the featured, but if we come down the bottom here, browse by domain. Uh, and

for me, of course, I jump into the maths one, see what's going on there.

And I mean, it says maths, but it includes stats as well. There's a power analysis calculator, stratified sampling, calculus solver, matrix calculator. So, there's a whole lot of

calculator. So, there's a whole lot of different AI powered tools in here in the agent gallery. And this is worth noting because not only can we get the

agents to do literature search, but we can pass we can we can kind of string commands together. So we can get them to

commands together. So we can get them to go and look at the some literature, find out some stuff, carry that stuff over into the next thing into the next thing into the next thing. So being aware of

some of the things that we know the agents can definitely do because they've been programmed to do like a power analysis handy for us to know about.

Okay, coming back to the side here. So,

that was the agent gallery. Um, there's

an AI rider. It'll help you with AI.

There's chat with PDF. So, we can put our own PDFs in there, but we can also when we search up a PDF uh for a literature search, uh, we can interact with the PDF. And that's going to be

really, really handy if you're reading very technical documents and really handy for those of you who are undergrad and PhD and maybe more newer to your domain. uh where it it's kind of like

domain. uh where it it's kind of like having a tutor or a helper that can just explain stuff to you as you go. So, it's

really valuable. Uh we've got literature review. We'll come back and look at that

review. We'll come back and look at that in detail in a second. Uh find topics is like a brainstorming tool. Uh we got citation generator, paraphraser, uh

extract data just we can throw a PDF in there uh and it will pull information out uh and an AI detector. So, let's

jump into literature review. And so we can do literature reviews in this one that says literature review. And we can also do them up in the new chat or up in the uh agents area.

So in our literature review, we can see we've got a text box and if I click that, it suggests some possible searches and I can do a standard highquality or

deep review. And I've already cooked

deep review. And I've already cooked some. So the deep review in particular

some. So the deep review in particular uh takes a little while to do. So

depending on how complex kind of independent if you're getting agents to do it certainly a couple of minutes in here maybe even a little bit longer sometimes if you get the agents doing something

fancy. So I've already cooked up three

fancy. So I've already cooked up three for us to have a look at and also to be able to contrast between the standard the high quality and the deep review.

uh up here. Another thing that's quite cool is you can put in your research question, but if you kind of you you know the kind of thing but not quite

sure. So if I was like uh creatine

sure. So if I was like uh creatine muscle something, it it'll take whatever you're putting.

So if you're putting keywords in there, it will turn those into possible research. So there's three pretty good

research. So there's three pretty good suggestions just based on the words that I put in there. So really really helpful. But let's go and have a look at

helpful. But let's go and have a look at the ones that I've done. So we'll start with the standard.

And so the standard it gives us a short summary.

Uh it references five papers in our short summary. And then it gives us a

short summary. And then it gives us a table of 10 papers.

And there's some filters that we can adjust to go and look at these if we like. Um, we can save the summary to a

like. Um, we can save the summary to a notebook. Uh, we can export all of these

notebook. Uh, we can export all of these references out to a reference manager.

Uh, something if you're PhDing and you're still figuring out exactly your topic. Uh, we got this button here, find

topic. Uh, we got this button here, find topics. So, if I click find topics,

topics. So, if I click find topics, then it's going to go off. We might need to come back to it's going to have a think. So, my original question was the

think. So, my original question was the impact of microplastics. So, this was something that I was doing some research for the environmental charity I'm

involved with. And I can hit find topics

involved with. And I can hit find topics and it will go away. It'll look at what it's collected and then it will suggest either gaps in the research or possible

narrowed research questions which is really handy. Uh so that's what the

really handy. Uh so that's what the standard looks like. Uh if we come back to literature review and we jump from standard to high quality,

we can see that the summary got a bit longer. It now sites 10 papers

longer. It now sites 10 papers and then we've got the 10 papers all listed here. So we end up with a table

listed here. So we end up with a table of the papers. Uh and if we scroll off to the side, we can actually add a whole lot of summary columns about and so it'll tell us stuff about each paper.

And we can even create our own one. So

something I've done a few times is I might have something create a new column like sample size. And so it will pull the sample sizes out of all of the papers and I'll just have a column that

shows me that. Um but insights results uh and there's a bit more. We'll have a bit more of a deep dive of that in a second.

So that was the the high quality and for just knocking out something really quickly.

That's probably sometimes if if your boss came to you and said, "I need a onepage summary about the impact of microplastics on the environment." There

we go. Done. Um, I would still check and verify the sources. I would still read this carefully to make sure it didn't say any nonsense. Um, but it's generally pretty good.

In fact, I say pretty good, I mean very good, actually. [laughter]

good, actually. [laughter] Uh, it's the New Zealander in me. uh you

can definitely have the recording those you that are asking about recordings.

Okay, so then we'll jump into the deep review. Uh and so the deep review is

review. Uh and so the deep review is takes longer and up here I can actually click out and it'll show you the steps. So the big difference on the deep review is I asked

my initial question, it comes back and it asks clarifying questions. So I said what is the impact

questions. So I said what is the impact of microplastics on the environment? and

it said, "Oh, well, do you want to know about ecological impact, health implications, geographical focus? Uh

where where are you interested in mitigation?" So, it it asked me for

mitigation?" So, it it asked me for clarification so it could give me the best possible summary. Uh, and so I answered that and then it took off and

instead of using my admittedly pretty poor uh, and pretty vague prompt, it started writing its own more detailed

prompts based on what I asked plus my clarifying points and then it started uh, going through articles, searching.

It does then a whole lot of filtering as well.

So we'll hide those steps. So, it went through 1750 1,750 relevant p uh kind of papers that were in the right ballpark,

identified 412 that were relevant to what I was after. It then wrote the summary with the 50 most relevant. And

you can see that now if we pop this out, this is a serious summary. This, in

fact, for for if your boss says, "Give me a one pager." Look at this. This just

goes on and on and on which um including kind of tables with summarization and

grouping um like when you ask it to do a deep review it gives you a serious deep review. Um and so if you are wanting to

review. Um and so if you are wanting to dive deeper into something um this is just amazing so helpful.

Uh so that's the deep review. it now is really big uh and very detailed. And so

in my table, I've now got 20 papers. Uh

but it says 20 out of 412. So if I scroll down to the bottom, it limits it to 20. So the page doesn't go forever,

to 20. So the page doesn't go forever, but I can click load more papers and it will show me the rest.

Um add next step. So from this I can then give more um more commands and do more stuff. Um

okay the let's we'll leave the comment about the AI checkers to the end. Does it work in linguistics? Um, Dpack can probably

in linguistics? Um, Dpack can probably give you some some specific stats of numbers, but they have a very comprehensive access to

um just all sorts of jour research journals and databases. Um, I have not searched for any domain and found it lacking. So, it will absolutely have

lacking. So, it will absolutely have linguistics in there. Um, the person that is asking about the linguistics.

Okay. So, we've got these summaries. Uh,

and that's pretty cool. Um, here's one.

And if we click on get PDF, this is a fairly recent edition. So, even those of you that are um have used CIA for a bit, if you if you

haven't used it for a while, if I click on get PDF, it can fire off a request to the authors, but you can also download from a library. So if you are at a university

a library. So if you are at a university or a research institution uh that uses what's called lib key uh and so this is answering the person about the payw

walls. If I click download from library

walls. If I click download from library I can connect to my institution uh and there are [snorts] thousands

here. So any any regional ranked

here. So any any regional ranked university should be in here and if it's not you should be able to go to your library and say hey set up lib key. Um,

so once you have set up your lip key, you can access the ones that are behind a payw wall the same way you would through your university or your research

institute library. Um, so that's how it

institute library. Um, so that's how it deals with payw walls. But if you don't have that, you can get it to filter down

to open access as well. So if we uh go just open access or ones where you can access the PDF, you can filter down to those as well.

Um so super super handy. Those of you that have known the hard ways of doing all this, this is pretty amazing. For every

single one of these papers, you can get a summary. You can even get it as audio

a summary. You can even get it as audio as a podcast and then you can chat with it. All of them also have their DOI. So

it. All of them also have their DOI. So

the digital digital object. I forgot what the I is.

digital object. I forgot what the I is.

That's pretty terrible. Uh so if we click on our DOI here identifier, thank you very much. Digital

object identifier.

Um it takes me to my source. So I can click on any of those DOIs. It will tell me here is my journal article. Uh in

this particular case, it's got all the download and PDF. But to be able to verify that this is real, I click on that uh and

takes me to the article and so I can see that it is a real article. No of those none of those concerns with hallucinations.

Uh and thank you for the person that added the DOI definition in the chat.

Very helpful. Thank you. Uh so if we come back, we click on the DOI, it takes us there. If we click on the name of the

us there. If we click on the name of the article, it's going to pop up a new tab.

And so on our new tab, uh, we have podcast. We can access the PDF. We can

podcast. We can access the PDF. We can

save it. We can site it. If we come down, we can chat with it. So there's

some default things you can ask the paper or ask it the chatbot about the paper. We can also throw it over to the

paper. We can also throw it over to the agents as well. Uh, so here is lots of different things that I could get. I could get the agents to write a

get. I could get the agents to write a critical review of the paper, analyze the citations, uh, to find unadressed researcher. There's lots and lots of

researcher. There's lots and lots of stuff here. Prepare a practice quiz

stuff here. Prepare a practice quiz even. So, pretty amazing. Lots of stuff.

even. So, pretty amazing. Lots of stuff.

And then I keep scrolling down here is the PDF. And I can over on the chatbot

the PDF. And I can over on the chatbot on the right hand side, I can ask questions, but within the PDF, I can just select something. If there's

something that I don't know like what is I can select it click on explain text and I'll get an explanation. So for

anything that's technical if you are maybe studying a domain and every so often you get something with really complex maths uh then it will give you explanations.

Um, not only that, but if we click back in, uh, and so there is very detailed explanation of the

garbage patch. Uh, it can even do that

garbage patch. Uh, it can even do that for maths and for graphs. So, here's a graph. Graphs may be a little bit

graph. Graphs may be a little bit blurry, so fingers crossed it's helped.

If I click on this, I can drag across any equation or any graph and it will throw it into chat as an image and then it will give me an

explanation.

Um, and that again for those of you that are newer to the domain and and just need a little bit of help with explanations or even the you those of you that are more experienced and just every so often you get a paper

and someone's just gone berserk with with all the Greek letters and the algebra and it's like well can you just explain to me in in like normal language um then

really really helpful and so we can see there is interpretation of that graph So that is chatting with the paper super

super handy. Okay. So next up that was

super handy. Okay. So next up that was all within the regular literature view.

We'll come back to home and next up we are going to look at agents.

Um so you notice I've got some credits up here. When we use the AI agents it

up here. When we use the AI agents it does use credits. So it is a finite resource. the literature reviews that I

resource. the literature reviews that I showed you, I think the deep reviews are limited to a certain

certain amount. Um, but the

certain amount. Um, but the anything with the agents is [snorts] um is using credits. The rest the the basic standard and the high quality literature

review, you can just do as many as you like. Um, so with the agent, we scroll

like. Um, so with the agent, we scroll down. There's some examples of stuff you

down. There's some examples of stuff you can do. And so you can see there's all

can do. And so you can see there's all sorts of research related things here from deep review like what I showed you uh literature reviews, writing drafts,

making powerpoints, making infographics, making a PDF report, uh searching particular sites, using Python, all sorts of stuff. You can actually type

whatever you like into this box. Um, so

you you're not just limited to the stuff, but this stuff is here to help you. So I might click

you. So I might click review literature. And you can see that

review literature. And you can see that it puts in the start of the prompt and I tell it what I want the literature

review about and then if I want to filter them in a certain way. And then

once I've done that literature review, maybe I actually want that literature review to be turned into a PowerPoint presentation.

And so I can click on that and it's just going to say do a literature literatur about the stuff include papers that are less than 10 years old,

whatever, whatever. Um, then create a

whatever, whatever. Um, then create a PowerPoint summarizing the findings. And

you can string multiple bits of this together. I can add my own in here as

together. I can add my own in here as well. There is a little bit of learning

well. There is a little bit of learning and trial and error. Sometimes I've

asked it to do stuff and it was very literal with what I asked it to do. It didn't It was kind of um, but

to do. It didn't It was kind of um, but in general, it's pretty good and certainly off all of the the templated ones. Very, very good. So, I'm going to

ones. Very, very good. So, I'm going to show you two that I have already done.

Um, the agents do take a little while to run. So, that's that's again why I've

run. So, that's that's again why I've kind of pre-baked these. So again, I was doing more of the microplastics uh work for the charity

and I got it to do a literature review uh about a particular component and then I said do the literature review and then I attach and then as with the [snorts]

agent I said do the literature review and then turn the key findings into an infographic.

So here is my summary about microplastics and where they're originating and the ecological

impacts and the human impacts.

And here is the infographic. Now

I need to go and sit down or I need to probably particularly for this middle one get a biologist to talk to and make sure all of this is correct. Always need

to do your diligence. always need to check stuff's correct, particularly when we start passing data from kind of a a component to component to component. Um

the but I was able to string those two together. So I'm going to show you one

together. So I'm going to show you one more example where I actually put three together. And again, this is kind of

together. And again, this is kind of thinking about how a literature review for different people has different purposes as well. So those of you doing a PhD, you might imagine I'm going to do

a search and then I might get some summarization, but it's really helping me find the information, absorb the information, but I at the end of my PhD for you, for you to be a doctor of

something, you should be an expert in that thing. Um, so, uh, as there was a

that thing. Um, so, uh, as there was a comment that just went past a little while ago about, uh, it not replacing.

Um and so you definitely need to maintain that knowledge. So this next one again with microplastics. Uh so this

time I chained three things together. So

my chain and probably gets a bit messy if I try and scroll up to the original thing. So in my prompt for this one, I

thing. So in my prompt for this one, I said, "Okay, we've been looking at microplastics. They're environmental

microplastics. They're environmental concern.

uh do a deep literature review of mitigation strategies for microplastics.

Then so that was the first first component. Then pass that into the

component. Then pass that into the search of grants.gov and search for any grants that match to

the um to those mitigation strategies.

And then third step create a presentation for me to take to the board which has the grants that are possible. So that

was chaining three different things together. Uh, and so if we go back and

together. Uh, and so if we go back and we have a look, um, I now have these slides where we've

got these different, um, different, uh, microplastics removal things and then US federal grant programs that it's going to line up to. In fact, even

including like contact emails, size of awards deadlines all of that pulled for me from the agents.

Um, so and then key research sources.

It's got some references in there as well. Uh, and again, I before I

well. Uh, and again, I before I presented that to the board, I would go and I would check all of these are correct. Um, it is it is my reputation

correct. Um, it is it is my reputation that's on the line with whatever I produce as a researcher and an academic.

Um, but how much time did that save me by doing the the literature search, feeding that into checking the grants that match up to it, feeding that into

the draft of my presentation? So, super

super handy stuff. So, all of that happens within the agents um to give you a feel for how many credits. So the

credit use uh depends on how how big and how complex the thing is you ask it to do. Um I think it was like 80 83 credits

do. Um I think it was like 80 83 credits just slightly over 80 credits for that three-step with a deep review

the searching of the go uh the grants.gov and producing the um producing the slides on top of it. Um,

so whichever of the plans, if you have a,000 or 5,000, it's going to you're going to be able to do plenty.

Okay, so we've got tons and tons of questions. What I might do,

questions. What I might do, let's have a quick scroll through and just see if there's any that are immediately about what I have on the screen before I go back to the slides.

Cool. Now I think what we will do is we will take you away. Uh we'll come back to the slides. I'll talk a little bit about ethics and integrity. Uh and

then I think we will finish off the Q&A Q&A at the end. Um ah excellent Joe Shanfu. Excellent thing. That was the

Shanfu. Excellent thing. That was the one thing I forgot.

Up the top here, connect apps and import data. And if we click on that,

data. And if we click on that, Zetero, Mendlay, One Drive, GitHub, Notion, and Slack coming soon. Um, all

of those you can integrate. So if you have a Zero library, you can connect it and then you can access it within Zpace.

So yes, good spotting and thank you for highlighting that. That was the one

highlighting that. That was the one thing that somehow escaped my brain.

Okay, there is an excellent question there about echo chambers that I think is overlapping into one of these. So,

ethics and integrity. Uh, some of this I've kind of been bashing on about throughout this whole talk because it is really important to me.

So some things that all of you should know particularly if you're a student but as an academic as well. Absolutely

know your institution's AI policies.

What are the policies on education on research on all the work that you're doing. Anywhere where you are producing

doing. Anywhere where you are producing a publication and submitting to a journal, you should know their policies as well. So that is absolute baseline.

as well. So that is absolute baseline.

Uh know what the rules of your institution are cuz otherwise you might get yourself into trouble. Beyond this,

think what is appropriate for your particular task. I've seen a number of

particular task. I've seen a number of people asking about AI detectors. Um, AI

detectors are unreliable. They are um discriminatory. Most places are stopping

discriminatory. Most places are stopping to use them. Some places still do use them. Um but if you are producing work

them. Um but if you are producing work and you are trying to skirt a detector because you know you are doing something wrong um you are going to get caught out at some

point either by an accidental copy and paste or putting something that is wrong and a particularly stupid into a publication or at some stage you're going to get found out because someone

is going to ask you a question and you over won't know an answer. And I've I've even even prei I've come across some people like that where they they worked unethically through their career and

then they got found out at some point and then that was the end of them. So I

would absolutely discourage any behavior where you go I need to make sure people don't know this because I'm cheating.

Just don't do that. Think what is appropriate for the task that you're doing. And so I gave some examples

doing. And so I gave some examples earlier on about for a PhD there's going to be a lot more of you and you absorbing information than if you are doing something that is just like a

simple fact check. Um always check anything and verify anything produced by AI. It's really important hopefully

AI. It's really important hopefully fairly obvious. Um be very mindful of

fairly obvious. Um be very mindful of anything you upload to AI systems. There's privacy implications, there's ethics implications, there's IP implications.

Um again coming back to the that comment about uh people in detectors really think carefully about what's your contribution to your discipline. If if

you are a copy and paste machine why would I hire you? Um so information is important and nuance and detail

in particular. So my next point AI can

in particular. So my next point AI can miss that nuance. That's where you as the subject expert are going to ensure at least for the next little bit that you're not replaceable by AI.

Um, second to last point there, the thematic blindness or algorithm bias. Be

really careful. Algorithms do have a tendency to kind of converge on a common thing. Uh, so there was a question

thing. Uh, so there was a question there, the echo chamber effect. Uh, DQP.

um you do need to be very careful and make sure you're kind of looking looking outside at references and other other perspectives in in your domain. Um last

one, you should be acknowledging where AI has been used in your work. So some

journals um and in fact I did have an example here is an example of an article that I wrote. There's me. Um so I wrote an

wrote. There's me. Um so I wrote an article about uh AI and medical education and at the end of the article we can see that this journal has uh this

declaration. So when I submitted I had

declaration. So when I submitted I had to make a declaration uh about how I used AI uh and then very key thing here

I accepted full responsibility for the content uh and then the details were declared to the editors. So for this article and scroll up to the top of the

article was about AI seemed reasonable to use some AI. One of the two things that I used was sci-pace. So I used sci-pace for some of my literature

review. I wrote my own literature review

review. I wrote my own literature review by hand myself but it was strongly informed by all of the papers that sci-pace helped me to find and I

declared that. So that was declared up

declared that. So that was declared up front. Uh and then I used chat GPT just

front. Uh and then I used chat GPT just for a little bit of brainstorming and ideation and again just was very transparent and made sure that I told

uh told the editors that's what I had done.

So all of this really really important.

Um if you want to continue to have a career in research and academia now that there are these amazing tools um you having that

contribution you demonstrating ethics and integrity are going to be really important.