top of page
Writer's pictureTina Gallico

How AI is helping Edaith: Generative AI and knowledge work

Updated: Nov 24

Generative AI tools have been indispensable for some tasks, unfit for others

 

Edaith and AI


I'm currently developing the first instalment of a collection of upskilling guides that cover the skills needed by workforces according to national authorities, including Australian Skills Organisation and equivalent bodies in the U.S., U.K. and Europe.


Using AI every day I've found it to be indispensable for some tasks, but I still feel as though I am better off without it for many things. In the explanations below, I gave a rating for each of my current major workflows, with 0 meaning AI is no help, and 5 being the top score.


NOTE: I'm experimenting with several tools simultaneously, so when I refer to 'AI' or 'chatbots' in the commentary it's shorthand for the following generative AI platforms:

  • Gemini by Google

  • Claude by Anthropic

  • ChatGPT by Open AI

  • Perplexity

  • Elicit

  • Jasper


Likewise for 'AI image generation':

  • DALL-E by Open AI

  • Adobe Firefly

  • Canva Magic Studio


The workflows


Strategy

0/5 For identifying the focus for Edaith resources, and what to research.


Chatbots have been no help. If anything their regurgitation of long lists of considerations have been a distraction. Best insights for Edaith have resulted from research and policy whitepapers, speaking with people, reading books, and selected podcasts. AI responses in this respect are broad and generic, whereas I need to be targeted and reflective of insights from experiences and feedback from people I aim to serve.


Research

2.5/5 As a starting point for finding the 'best' information.


AI tools give a little initial boost each time I start a new topic, but to achieve the quality of work that I'm aiming for, I need to go much deeper into the subject matter and be much more selective in terms of identifying what's worthy to write about. Even when I specifically ask for academic or scientific research, the generative AI tools always brings back blog and website articles that generally aren't research backed.


The best AI research tools I've found are Perplexity and Elicit. I'm mostly using Perplexity, which has pointed me in the right direction many times, but can only be an initial check due to research paper links often being broken or unretrievable, and the sources commonly being blogs and a mix of websites, some reputable e.g. university faculties and programs, some not. This week Perplexity added an option to subtract certain sources from the sources cited, but doing this does not result in a reiteration of the references with new, more credible sources.


An example of an output by Perplexity AI
An example of an output by Perplexity AI

I haven't been using the chatbots for summarising specific research such as journal articles, thesis publications, reports or books because when I do it rarely pulls the information that I would choose from the source, or doesn't interpret the findings in a way that is useful for my purposes. I know there are more detailed prompts I could try but I don't see the point in this as the content isn't anything I can't understand with some effort, so I'd rather spend the time and be sure I'm identifying insights that I think are the 'best' takeaways. This is definitely is not the easy path, but I'm happy with the results and so will persevere.


When I'm reading for specific topics I discover related information that is important to include in the broader work, even though I wasn't specifically searching for it at that moment. Also, delving into topics increases my exposure to different ways information has been organised, helping to improve my ability to conceptualise taxonomies and information design, which is a key aspect of the Edaith resources.


For me, chatbot dialogue is a scoping tool for relevant themes identification, then I make a judgement about what is key information working with more reliable and peer reviewed sources via Google Scholar, arXiv, research institutes, and using the academic search databases when I'm working from a library e.g. Taylor & Francis, JSTOR, SpringerLink, as well as books, to decide what should be included.


Outreach communications

3/5 For first drafts of email correspondence and advertisements.


Whilst I still undertake background research for outreach and am very specific with the ask in any email correspondence or public messaging, I've found generative AI to be put well to use here. With the prompt I list the specific things I'd like addressed, often the output includes verbatim items from my list, but overall I get a draft structured text that I can then edit for use.


Learning

4/5 For explaining a specific extract of information in an alternate way.


Chatbots are a good resource for learning support. For example, I'd been interpreting a diagram and wasn't sure about some aspects of it. I took a screenshot and ChatGPT's explanation did the trick.


Learning assistance with ChatGPT
Learning assistance with ChatGPT

General writing

3/5 Only for writing that is universal, without nuance, or a point of view e.g. basic 'how to' instructions.


The AI chatbots I've been using aren't hitting the spot in terms of making judgements about what is essential information, and what is periphery information. Also, I prefer to categorise information in different ways to their structures, or make different connections between topics.


AI has been helpful to get started with first drafts of ubiquitous information, such as how to use some of the methods I'm including in the guides. The outputs need additional information added, are far too long winded and must be highly edited down, and sometimes I don't refer to them at all. But overall, considering the volume of work I'm undertaking, I'd say using AI dialogue as the starting point with the method instructions is how it's been most helpful to speed up my process.


Design, diagrams and imagery

0/5 Not useful for my purposes


I have experimented with AI image generation and will not be using it for now. The images are either nowhere near the aesthetic and quality I am aiming for, look fake due to incongruent elements, or have a strange AI generated vibe (IMHO).


As I'm a fan of simple, geometric graphic design and able to work with Adobe Creative Suite, so I am using Illustrator, InDesign and Fresco to create diagrams and design visual elements to support the written work. I would love to work with a professional graphic designer for this workflow when budget allows.


Overall


The process for my work is mentally challenging, expansive and time consuming. Truthfully, for the first draft of this piece the title was:


The slow, frustrating, but necessary grind of creating.


But AI is helping. Generative AI platforms are like having 24/7 assistants with suggestions on ways to move forward, but they can't yet deliver the critical work if your goal is to create something exceptional. Overall the impact does not feel like my job is easier, although certain aspects of it are definitely faster to get off the ground.


Tina


 


 

Get your knowledge haul



Comments


Commenting has been turned off.
bottom of page