Dr. jonny phd on Nostr: nprofile1q…hcvwk I think of it like different layers for different reader ...
nprofile1qy2hwumn8ghj7un9d3shjtnddaehgu3wwp6kyqpq9d9p04u4xfysdy92fycw947jrca3xve2gnsauysshzewxvmz8dmsxhcvwk (nprofile…cvwk)
I think of it like different layers for different reader intentions. We already communicate the same argument a bunch of different overlapping ways, and a reader will seek out different ways depending on their intention - read the abstract to get the gist, read the text to get the argument and context, check the methods, see the data in a plot, pursue the raw data, etc. No single modality allows for all intentions, so a description of the shape of data and what's in the plot won't let someone independently fact check you, that's true, but plots usually don't do that anyway.
Put another way, the "honesty" of a plot is just a reflection of the honesty of the plotter - there are enough ways to plot data misleadingly that support the argument you want even if the data doesn't actually do that. Someone who is very interested in people getting a rich sense of the data will show a full alpha shaded scatter plot along with the regression fit and error ribbon from a model with some sophisticated error accounting, but someone who doesn't care about their reader being able to check them (or idk they use excel, never thought about viz, whatever) will just throw the mean on there as a bar chart, maybe with the standard error because it looks smaller, and call it a day.
You already said it - the raw data being available is another complementary layer that should be available, and you can do the best you can to make it tidy and accessible but you won't be able to solve accessible computing for ppl with vision disabilities with writing the right alt text, then it's sort of out of your hands re: whether there are the tools for someone to explore the data on their own. You could take a bunch of steps like sonifying the data, but at that point the unsighted person probably has their own techniques and tools they prefer and can handle it if they want to (or else the tools don't exist, and that's a whole different conversation re accessible computing)
Providing the interpretation basically is your whole role as the researcher, so I agree with what you're saying re: someone needs to be able to disagree with you and making that possible for as many ppl as possible is the goal, but i don't think that means there's no point in visualizing the data and then describing that viz, just that the plot is one part in an overlapping set of modalities that does a relatively circumscribed thing, to give the feel for the data and show some critical details like "is this line above that one, do they cross some important value and where, is the series smooth or are there sharp deviations, what's the error spread like" etc., and that intention for feel of data is what goes in plot alt text because that's what the plot is communicating to a sighted person that is distinct from a humongous table of raw values
I think of it like different layers for different reader intentions. We already communicate the same argument a bunch of different overlapping ways, and a reader will seek out different ways depending on their intention - read the abstract to get the gist, read the text to get the argument and context, check the methods, see the data in a plot, pursue the raw data, etc. No single modality allows for all intentions, so a description of the shape of data and what's in the plot won't let someone independently fact check you, that's true, but plots usually don't do that anyway.
Put another way, the "honesty" of a plot is just a reflection of the honesty of the plotter - there are enough ways to plot data misleadingly that support the argument you want even if the data doesn't actually do that. Someone who is very interested in people getting a rich sense of the data will show a full alpha shaded scatter plot along with the regression fit and error ribbon from a model with some sophisticated error accounting, but someone who doesn't care about their reader being able to check them (or idk they use excel, never thought about viz, whatever) will just throw the mean on there as a bar chart, maybe with the standard error because it looks smaller, and call it a day.
You already said it - the raw data being available is another complementary layer that should be available, and you can do the best you can to make it tidy and accessible but you won't be able to solve accessible computing for ppl with vision disabilities with writing the right alt text, then it's sort of out of your hands re: whether there are the tools for someone to explore the data on their own. You could take a bunch of steps like sonifying the data, but at that point the unsighted person probably has their own techniques and tools they prefer and can handle it if they want to (or else the tools don't exist, and that's a whole different conversation re accessible computing)
Providing the interpretation basically is your whole role as the researcher, so I agree with what you're saying re: someone needs to be able to disagree with you and making that possible for as many ppl as possible is the goal, but i don't think that means there's no point in visualizing the data and then describing that viz, just that the plot is one part in an overlapping set of modalities that does a relatively circumscribed thing, to give the feel for the data and show some critical details like "is this line above that one, do they cross some important value and where, is the series smooth or are there sharp deviations, what's the error spread like" etc., and that intention for feel of data is what goes in plot alt text because that's what the plot is communicating to a sighted person that is distinct from a humongous table of raw values