Thank you both, really useful.
I agree with ChatGPT being very unreliable with the truth

Have you come across scite? It’s supposed to be powerful in the domain of literature reviews.

Best,
Till

On 21 Mar 2025, at 14:50, Stewart Robertson <stewart.robertson@stir.ac.uk> wrote:

Afternoon Folks and happy Friday,

Just a brief AI update following our last session in November and ahead of any potential future AI lab session - things have moved on considerably. Here are my thoughts on current AI tools I've been dabbling with, along with my personal opinions. I'm happy to demonstrate any of these at a future lab meeting if there's interest.

      ClaudeAI

This is my default all-purpose general AI tool for everyday use. It has limitations regarding context window (the amount of PDF and other attachments you can feed it), however, I have the paid version of Claude and consider it the best tool for most of my AI tasks. It became even better a few months ago with the launch of Claude 3.7 Sonnet. I use it for:
  • Brainstorming and developing ideas - recently I've been using it to help develop vignettes I'm working on
  • Data analytics - feed it a CSV file and it will help analyse and develop the data, including adding new datapoints. I was working on presenting case studies relating to global corporate failures - it took the CSV file and was able to add extra data points like Country of Origin, Date/Year and highlight key words
  • Summarising information - you can get it to highlight key words in text and use colour coding to apply different types of sentiment. I did this recently for a data table of 10 cases of firms and individuals receiving regulatory fines - Claude highlighted key words and used colour coding to categorise different types of failure causes (e.g., Audit, Governance, Culture, Fraud, etc.)
  • Generating interactive HTML visuals from any PDF
  • Creating data tables from a PDF and getting Claude to data-scrape for you to create a dataset
  • Converting a PDF into a Word Document
  • Coding - this is by far the best coding co-pilot for any language, in my opinion! 🙂
  • Processing transcripts - either from Teams recordings or YouTube videos - it does a fantastic job of creating useful notes/summaries. It doesn't matter if the transcription failed - for some reason Claude is able to cope with mistranscriptions
      NotebookLM

  • Free version is excellent - up to 50 sources
  • Best tool by far for analysing multiple sources relating to any subject you're interested in - from PDFs to YouTube videos to basically any attachment you can think of
  • This tool is the best at controlling and managing so-called 'hallucinations'
  • Great tool for learning - now you can interact with the audio/podcast that it generates
  • This week, they introduced a new mind-map feature which I love - I created one on 'Fraud, Behaviour and Deception' which turned out fantastically. Unfortunately, it doesn't allow you to share the output yet, but I'm happy to demonstrate this (can be saved as a picture file though)  

      Non academic 'Deep Research'
              
            ChatGPT

  • ChatGPT - Deep Research Mode (I've been running this using the o1 model and NOT the default model)
  • I've been using this tool at work and was impressed
  • It can access Google Scholar and academic sources but doesn't do a great job academically due to all the paywalls that their agents can't bypass
  • Good for sentiment analysis based on recent media stories - you can instruct it to look at Reddit and other forums too
  • You need the paid version to access this - and due to the processing resources required, they have limited Plus users to 10 reports per billing period, so use them wisely

            Gemini (Google)

  • FREE!
  • Gemini was the first to introduce this - more recently they have enhanced its capabilities by using Gemini 2.0 (previously used 1.5)
  • Similar usage to ChatGPT - although Gemini was not nearly as good as ChatGPT on this. That said, I HAVE NOT tried it since it has been updated with their best LLM model, so expectations are that it should be significantly better

      Academic 'Deep Research'

  • Elicit - IT'S DEAR at $49 a month...  @Liz has also been dabbling with this tool as well.  I upgraded briefly to try it out but quickly cancelled my subscription.    Worth watching this video which outline how it does deep research and automated systematic reviews.  Elicit Reports — Deep Research for actual researchers

    I think this will get cheaper, but my advice would be to only use this if you have a specific and precise idea of what you want to get out of it. You can essentially give it your research question, and then it helps you refine various parameters for it to focus on when generating the report. You can see the one I generated on whistleblowing here.
                        
  • Scispace - alternative to Elicit - this MAY be much better than Elicit but I don't know - I haven't tried this one, but it can do similar things to Elicit (and more by the looks of things)

  • Research Rabbit - A tool for visualising connections between academic papers and researchers. It allows you to enter a paper or author of interest and then maps out related works in an intuitive network, showing how different research connects and evolves over time - particularly useful for literature reviews and discovering new relevant sources within your field.  I am aware of this tool but haven't really delved into it at all...

      Image Generation

  • This is where it gets fun - I was experimenting with these for my vignettes and ended up getting an annual subscription to Midjourney.   I was incredibly frustrated with ChatGPT around getting two images and then altering one - frankly, it was impossible with ChatGPT. Midjourney allows you to highlight a specific part of the image and then using prompts, you can alter the image with precision. The quality and capacity of this tool is mind-blowing. HOWEVER - it comes at a price - to get the specific functionality you need, they nudge you into taking out an ANNUAL subscription at £89 per year. So I'd only advise using this tool if you have a specific purpose/need in mind - a lot cheaper than hiring an artist... Happy to demonstrate this - just ask - also happy to try and generate images for anyone using my account.

  • Hot off the press - Gemini launched their latest image generation model the other week - and you can now feed it an image and change aspects of it using prompts. Watching some of the videos online, it looks amazing - however, I was less impressed when I tried it myself, but then again I've only dabbled a bit with it since launch. Access to this is free with a Google account, but through something called GoogleAIStudio - you then have to select "Gemini2.0 Flash (image Generation) experimental" as the model to experiment with

      Power Point - Information Art and diagrams
                                         
  • Napkin - It's free and simple to use - basically can turn simple lists or process narratives into nice AI-generated infographics. I've used this recently on two PowerPoint decks I've been working on - and it just gets better...
                  
     Consensus

  • This is a good one for fact checking assertions folk make - @Katie - one for you - I Asked it "Whether the moon landings were fake?"  and got this response (thought it did a pretty good job):


    • The question of whether the moon landings were faked has been a topic of conspiracy theories for decades. Despite overwhelming evidence supporting the authenticity of the Apollo missions, a minority of people continue to believe they were staged.

      Evidence Supporting the Moon Landings

    • Scientific and Physical Evidence: The Apollo missions left behind substantial evidence, including radio transmissions, photographs, film, and rock and soil samples from the lunar surface, which have been extensively analyzed and verified by scientists worldwide7.
    • Historical and Technological Achievements: The Apollo program is recognized as a significant technological achievement, with twelve astronauts exploring the lunar surface between 1969 and 19729.
    • Visual and Topographic Analysis: While some conspiracy theorists claim that images from the Apollo missions were manipulated, topographic analyses and comparisons with simulated lunar views have been conducted to verify the authenticity of these images8.

      Conspiracy Theories and Public Perception

    • Conspiracy Theories: A small but vocal group has long claimed that the moon landings were faked, often citing supposed photographic anomalies and government distrust as evidence1 4 9.
    • Public Belief: Opinion polls have shown that a minority of the population, ranging from 10-25%, have expressed doubts about the moon landings, influenced by conspiracy theories and media portrayals2 4.
    • Psychological and Sociological Factors: Belief in such conspiracies is often fueled by pre-existing prejudices and a tendency to seek information that confirms these beliefs, a phenomenon known as confirmation bias1 5.

      Conclusion

      The moon landings are supported by a wealth of scientific evidence and are considered one of humanity's greatest achievements. While conspiracy theories persist, they are largely based on misinformation and a distrust of governmental and scientific institutions. The evidence overwhelmingly supports the reality of the Apollo missions.
                  
Anyhow - I wouldn't have written this email if I wasn't stuck at home with my leg up. Hopefully, it will be a useful resource and provides us with some options for demonstrations at a future lab should there be interest.

Cheers,
Stewart

PS There are LOADS more tools out there -  one that many people use is Perplexity - I haven't tried this tool, but it's mostly free, and I always hear good things about it...


Stewart Robertson
Behavioural Science PhD Student
Stirling Management School
University of Stirling
Stirling
FK9 4LA

[in]   [🌐]   [✉️]

Scotland’s University for Sporting Excellence
The University of Stirling is a charity registered in Scotland, number SC 011159
--
PhD-BehSci mailing list -- phd-behsci@lists.stir.ac.uk
To unsubscribe send an email to phd-behsci-leave@lists.stir.ac.uk


Scotland’s University for Sporting Excellence
The University of Stirling is a charity registered in Scotland, number SC 011159