And while I have been trained not to draw broad consequential claims from studies whose external validity may be limited, applying this insight or keeping in mind the principle of compromise in policymaking and peacebuilding activism can be the answer we are looking for in a polarized world of today

One of the most crucial findings from over four years of my research in Leiden is that compromise is a highly valued principle among voters (at least in the domain of legislative politics)

Thank you so much, Babak!

this implication for political communication and polarization in social media (and beyond) is explicitly reflected in #grok's instructions, and I can only give kudos to the developers, while hoping that this situation will not change in future.
And that's all from me on this thread 🤓

...but the main output of that #grok analysis for me was that, contrary to my expectation, this AI has not passed judgement on the involved people and institutions in a way you would expect from the technology developed by the company motivated and dressed in a particular party/ideology colors

now, let's go to case 2. There, I replicated the above steps but applied them to the series of controversial posts related to an 'ophef' of 2022. As that scandal is no longer relevant and has been resolved, we will not look into the details...

looking at these 'thoughts' I would argue that we can use AI tools in education, even if only to demonstrate to students how one can organize their interaction with resources, opinions and judgements. Of course, as with everything, the success of such approach would depend on students, but still

it even browsed polisci research papers to see if the hoax has been widely discussed in the discipline (I would be cautious to trust that this analysis has been exhaustive, similarly to how we should not trust ChatGPT's essays, yet it gives me a good overview and start point for my own research)

#grok 's extended output allows one to check what goes into its analysis process: thus, I saw the key points, as well as the summary of the whole debate, relevance of it to my own background and field, broader systemic issues related to the publishing industry and academia, its implications etc

...however, being able to conclude this from "[it] has blown my mind but raised many questions" is quite impressive to me.

but I was surprised, and, threading carefully, I would even say - positively: #grok inferred the range of the arguments I was thinking about fairly accurately. Now, of course, this is not rocket science, and I do not presume to have cutting edge opinions on the topic...

and here I was very curious, considering the team and the person behind #grok, I was ready for anything...

however, it asks itself an interesting question: "I am wondering what specific concerns Elina had". For context, see the post on the sokal squared hoax below. So, naturally, my next step was the following:

case 1: I did not give #grok any prompts, just pressed on the the first post of the identical two-comment thread published on BSKY and X (image below). The AI's 'thought process', available to the user, shows that #grok is unsure what exactly the user wants in this scenario, a content summary or?

partly, this is the case because #grok has a good set of instructions it keeps referring back to, it almost looks like human logical reasoning process. How do I know this? I used a couple of my own X posts, controversial and not, to see what kind of interpretation and analysis can this AI perform

first of all, good news, the AI still cannot read our minds, but it comes pretty close to asking itself 'why questions' as per Judea Pearl's and Dana Mackenzie's brilliant book (2018)