### Chatbots and some more scientific alternatives Language models have recently revealed a huge variety of novel entries into (socio)linguistic interaction. From simple information-querying which aggregates more sources than we would normally be able to read; to translation between languages, disciplines and even ideologies; to the exploration of social dynamics and emotional engagement with interfaces which are far from “human.” None of these things come without their problems, of course. A particularly pressing problem, which combines key issues of bias, opacity, monodimensionality and explainability, is the chatbot interface we have mindlessly accepted as the sole way in which we interact with these systems. My main line of investigation for the fellowship, and what I would like to share with the CLI community and receive feedback on is the following: what possibilities exist, beyond the chatbot model, that can assist 1) academic research in novel ways, and; 2) that can present (possibly also to the wider public) the complex architecture hidden behind the chatbot interface, so as to promote a more nuanced vision of their outputs? Regarding 1) the idea here would be to improve scientific research by designing new tools that deal with language and the complexity of our theories, and regarding 2) the idea would be to lessen the black-box myth and the commercially-driven AI hype. The presentation style below is inspired by [Felienne Hermans](felienne.nl) (who handwrites her presentations, and writes critically, consciously and comically about AI and related topics).       ![[Untitled_Artwork 5.jpg]]         ![[Untitled_Artwork 4.jpg]]         ![[Untitled_Artwork 3.jpg]]         ![[Untitled_Artwork 2.jpg]]         ![[Untitled_Artwork 1.jpg]]         ![[tech in academia.jpg]]         ![[Untitled_Artwork.jpg]] %% Present, in 4 slides or 5: some things we take for granted: 1 Breath (masterman), 2 arrows (own work), 3 chatbot interface (eg debate mapping): 4 we can do more! https://www.nngroup.com/articles/ai-paradigm/ : “AI is introducing the third user-interface paradigm in computing history, shifting to a new interaction mechanism where users tell the computer what they want, not how to do it — thus reversing the locus of control.” Ideally: UX changes as user desires?! But we want repetition/reliability... differences: - from **sequential logic** (A then B) to **spatial/structural logic** (A relates to B). we can see, even if we do not end up going for them, the distributions of OTHER paths not taken, domains not inquired into, etc. - society library inspiration: _Claim, Evidence, Counter-argument, Context._. Result, not "ask" the AI; you "audit" the AI. You can physically see the branches of logic where the AI might be hallucinating or where the evidence is thin. - from prompting to spatial organization of thoughts (we are already using Miro a lot, and it is confusing, what about an actually effective, dynamic miro?) - from character to **complex machine** (that is what it is, why do we want to lie to ourselves:?) sliders for "Temperature," "Top-P," "Frequency Penalty," and "Attention Weights." This is essentially a GUI built on top of the LLM’s inference parameters. You can use **Gradio** or **Streamlit** to create these control surfaces. Important: removes the "personhood" of the AI. You aren't asking a friend for a favor; you are tuning an engine for a specific performance. - taking you to the right part of a document, highlighting, not talking to you examples: from medical domains: outliers: huge problem (i am one) sociology lingo / philosophy lingo economics and sustainability etc. ### Notes from meeting - AI @ EUR: problems of connecting absolutely everyone - governance: how? Updates: how? - Strategic goals are: to become influencer in responsible ai, empower students and staff in ai competencies, mobilize ai as pedagogical catalyst in teaching/learning, integrate ai into “DNA”, make experimentation a strategic asset - idea: create the same tools and frameworks for the entire university ### Jop Dispa - check lavender (israel) - check his text in dutch -