AI Assistance

Possibilities of AI

At the time of posting this topic (Aug-2026), generative AI-assistants (like ChatGPT, Claude and Gemini) have limitations in making relevant judgements about personal and social functioning.

However, because an AI-assistant can read very quickly and hold much more in memory than a person can, it can assist me in connecting with material found in the more than 3000 webpages on this website.

AI-assistants, working closely with me, definitely increase my efficiency, especially in addressing cross-framework puzzles and issues, and in entering new areas of inquiry.

Help falls into three categories as elaborated below:

  • as a sparring partner when investigating new frameworks;

  • as a proofreader;

  • as an unbiased reviewer of existing published THEE frameworks.

AI as Sparring Partner

I found that an AI-assistant can help me with checking emerging propositions against previous formulations. While I can do such work myself, checking past frameworks is extremely time-consuming. Also, it is very easy for me to make errors of both omission and commission.

Because a taxonomic approach in human functioning is largely absent in the literature, AI assistants, however advanced, lack taxonomic sophistication. None has any background in this particular taxonomy (THEE), nor any feel for what it is be human. So they rarely produce correct formulations.

I can always draw on my own experiences and intuitions, and I use themm to explain to the AI-assistant why their suggestions are incorrect. I try to give guidance based on taxonomic principles that are well-established. Doing this stimulates the AI-assistant and, more importantly, gets my own mind working along the right track.

AI-assistants use language in a way that is repetitive and not in character with this website. So it was never sensible to copy and paste chunks of text. More seriously, the AI-assistant repeatedly made mistakes or interpolated inappropriate assumptions. When I confronted and explained, adjustments followed but old wrong thinking frequently found its way back into responses. Vigilance and my sense of responsibility reduce the risk of such errors influencing conclusions.

AI as a Proofreader

I do not expect an AI-assistant to copy-edit. However, it can proof-read and in early 2026, I started using it for this purpose.

It tirelessly identifies typos, diverse grammatical errors, extra or duplicated words, garbled sentences, wrong spelling, inappropriate punctuation, mismatches between text, tables and graphics, stray spaces, inconsistent layouts, obsolete THEE-names, and obvious formulae errors.

The AI does not act as an agent. It presents me with a list of issues and needed corrections, an then d I check them one by one and alter the topic manually.

At the present time (October-2026), about 40% of the website has been proofed.

AI helps Overcome my Bias

I used an AI-assistant to check on the psychosocial pressures in structural hierarchies developed prior to establishment of that conjecture.

The initial goal in this study was to determine whether the descriptions of groups and groupings in structural hierarchies support proposed assignments of 1° and 2° psychosocial pressures based on just one recent case. This would allow me to assert those assignments as a general principle.

Because many frameworks had been developed prior to conceptualizing psychosocial pressures, finding their presence implicit in the formulations would offer some validation.

If I had done this review myself, I would have been biased. The inevitably limited understanding of the AI-assistant was a worthwhile price to pay for its lack of prior commitment to my conjectures. While AI-assistants are programmed to agree with users, and so may possibly seek to please me by supporting my conjectures, intelligent arguments with extracts from the older frameworks offset this to some degree.

As authors invariably say about advice from colleagues: any mistakes that remain in this study are my sole responsibility.

Originally posted: 15-Aug-2026. Last amended: 31-Oct-2026.