Practising Programming in C

Notes

  1. Decontextualisation-unfriendly communication system
    • Tackling sourcelessness, e.g. this tweet
    • Providing peripheral vision, e.g. this tweet
    • Record history of...
      • Message edits / deletions
        Deletions themselves only hiding the message from the public. It still exists for the author / admins to view
      • Profile changes – e.g. avatars, names, descriptions – assuming we have profiles at all, which we may not
        This was prompted by the Tory press office rebranding itself on Twitter
    • Contexts
      • Language
      • Temporal
    • Referencing...
      • Subsections of messages
      • Multiple (subsections of) messages at once
      • A set version of messages, defaulting to the version current at the time of writing the present message
      • Profile description(?)
      • Specifying which subsection(s) of our message applies the reference
      • Application of references after the fact, by anyone, perhaps with the consent of the writer and referee
      • Description of reference, e.g. direct Q&A, acknowledgement, restatement, directing attention
      • Displaying referenced message inline, especially if external to our system
      • Highlighting of key points within reference
      • A kind of specificity "score" that readers may give to mark "the extent to which the specificity of the reference aligns with the (subsection of the) message's content", inspired by this question from Jenny Andrew
    • Analysis of message while composing, and auto-searching existing messages for a possible reference
    • Split view: user defined / auto, to view searched messages
    • Truthfulness report for existing messages, inspired by this advocacy of Joan Donovan
      • Indicator, possibly traffic lights-style, e.g. unverified, verified true and verified untrue
      • Measure of the amount of "run-off" or work that has gone into verifying / debunking the message
    • Description of behaviour (ongoing analysis required)
      1. Purposes
        1. Aid moderation
        2. Clue people in to the kinds of messages and people they're engaging with

          A sub-purpose of this one is to try and let people assume better (than the worst) about "strangers". A case in point being this short twitter exchange from 26th July 2020. Initially, all I wanted to reply with was "Don't we want poverty to die, though?" But I knew that the person didn't know me – as I didn't know they existed until the day before – so put in the two agreements to try and clue them in to the idea that I:

          1. seek cooperation / agreement, not fights, and
          2. was understanding the spirit of the point and replying in good faith, while noting a possible mistake, a possibly unintended reading of the message (it happens, as I've had myself, to the enjoyment of both of us!)

          Clearly this didn't work, and the brief conversation proceeded as between people not on the same page, with me sort of doing the "seeking to prove my social conscience credentials" thing (in an effort to spell out my agreement, since it hadn't gotten through the first time), and them doing the "rebuffing smartass reply guy-type" thing.

          Without this feature, we can end up doing this "proving" / "rebuffing" dance all over the place. With it, hopefully we may be freed up to progress our conversations in more productive directions, and engage in more enjoyable ones, that we enter with a fair idea of what we may get out of them.

      2. Possible approaches
        • Compute the proportion of Nouns, Verbs, Adjectives and Adverbs, to see how messages correlate with genres
        • Analyse mood, i.e. imperative, interrogative or declarative
        • Word embedding (maybe?)
      3. Contexts, from most local to most global
        • message-local, possibly subdivided into smaller linguistic units
        • messages that reference and are referenced by the local message
        • conversation-wide
          Noting that conversations themselves may require some work to delineate, if referencing is used as extensively as we intend
        • entire profile's output
        • entire profiles' output of all in conversation
        • entire community's output
    • Categorisation, automatically suggested once possible (e.g. above some confidence level)
    • Maybe some sort of permissions, e.g. groups of users able to edit the same message
    • Refer people to messages, without having to emit a full-blown message yourself
    • Schedule for later, from this post by Elsie

    This article makes points that we aim to address (from this tweet)

  2. Political / Social Development Classification Framework (tweet)
    Maybe inform James Mitchinson and Samantha Harman, paper editors interested in fighting back against disinformation
    • Brexit-specific version
      • EU compatibility of legislation: could it have been legislated without Brexit?
      • Brexit-promise compatibility of development: did Vote Leave campaign promise it?
      • Impact assessment – is it a net gain / loss for the people directly involved or, say, 95% of the population – and could Brexit be plausibly credited / blamed for this impact?
      • From Margaery Gerrard (tweet): "Why don't we identify key success criteria? Not many, maybe 6 factors that can provide a measure of how successful brexit is in improving
        1. the economy
        2. health and welfare of citizens"