There are increasing efforts to deal with misinformation and disinformation on the internet but this is part of a larger issue with the hugely saturated sphere of internet content that often lacks context and deprioritises clarity and understanding. How do we properly balance safety and autonomy, (satire and truth, mystery and clarity) and most importantly who gets to determine this? More recently there have been some projects addressing provenance in the sort of ways I’ve been hoping for. Here are some more considerations on context.
- “Educating people about how to assess risk is obviously a worthwhile exercise. However, lecturing them about what they should and should not fear does little to improve matters” - How Fear Works (Frank Furedi)(but i didn’t agree with enough of this book to bother finishing)
- How to balance children’s rights, child safety, child autonomy
- Demystify the internet without ‘shielding’ kids from stuff, limited/mediated access from a third-party
- Maybe a sidebar of explanations, context, guidelines, and someone to talk to who is actually trained to chat with kids for education, triaging queries etc
- ‘informed consent’ model of content warnings rather than ‘safe search’ style or ‘trigger’-only warnings
- Some AI models providing age-stratified provision for children - how up to date are ‘reading levels’? Is age still used for this — i looked this up where is it
- Is there anything that actually ethically and accurately screens for ages? I doubt it. But also proven issues with self declared ages. But if no direct interaction between kids of those ages, maybe self declaration is fine?
- Unknown/unofficial/anon/disconnected authors? ethics? right to be forgotten, right to be uncontactable, right to disown or defocus certain past works and frame your past in your own way. Link to the advice for citing trans academics
- identifying and indicating intent, direct interest, personal opinion, third-party viewpoints, many are subjective links to expertise
- Some intentions of content
- To entertain, perform
- To garner attention regardless of impact, to bait, to troll
- To educate, inform, share knowledge, organise
- To misinform, incite hatred
- To persuade, convert, sell, convince
- To sell based on false promises, unsourced claims
- To constructively criticise, discuss, clarify, correct, add on, boost others’ content
- To praise and congratulate
- To share, reflect, foster community, ask advice, be vulnerable
- To inflame, incense, criticise with call to action, boycott
- To support bigots and incite hatred, to target the vulnerable
- Categorisation generation, content taxonomy, granularity, scope
- fact checking AI is v questionable
- Content provenance, authors, publishers, natural language descriptions of trust, trust mechanisms
- Browser extensions or standalone browsers that value provenance more in visual hierarchy, ‘reader mode’. May have additional security features, translations, summaries?
- Open hardware implementation (or linux browser for kindle-like devices) of content-viewing device- like an e-ink e-reader but also with readable internet content
- Generating plenary questions from literature on e-readers
- AI generated summaries/fact-checking/provenance on video content through transcripts, ethics
- clearer forms of self-directed tone indication/intent declaration in micro-content to assist with social misunderstandings?
— TV tropes, know your meme, urban dictionary where it isn’t just hyperlocal, clickbait headline summarisers etc, combined and expanded to other public deceptions, jokes, common patterns, social phenomena etc for various cultures. General language AI can explain phrases or jokes I don’t have the context for but community-comment-twitter style is better. ethical considerations but maybe use to general demystifying of social patterns
- since then, many developments on individual platforms as well as large increase in uptake of AI models since ChatGPT came into public use, making this idea more mainstream, providing better examples
- providing provenance to articles way easier than interactive sites and micro-content
- on making sources very clear to allow informed consumption
- time relevant comments - eg some stuff I comment about a game update is only relevant if you see it within eg first week of update being out
- finally some projects are focusing on proper provenance!
- but also many v questionable ai projects on this
- which context is relevant when
- many of my thoughts here have been altered by further understanding on trust in provenance - as trust is relational, context is required for recommendations etc
- intersubjective methods are the best we can do. context-dependent context. views.
Resources