Closed matthew-blackman closed 6 months ago
@samreid mentioned that an alternative approach would be to sonify the histogram from bottom to top. His reasoning is that as data is collected, the histogram forms from bottom to top. Ending with the data in the highest bar would also emphasize the mode bin.
As we implement this system, we should keep things as flexible as possible to compare different approaches and keep our options open for experimenting with sound mappings in the histogram.
@catherinecarter (via Slack):
I’d add, “from left to right” so it’s clear the focus is on how the shape builds from left to right. Also, I’ll admit that when I think of a histogram shape, I think of a Normal distribution as sounding like low to high to low, which would change the mapping as it’s currently being discussed since this would map the height rather than position on the x-axis. It may or may not be a good idea to do that, but I’m starting to picture the sound getting increasing higher, which may or may not lead the user to think it’s an exponential or square root function. I found this: https://www.dailymotion.com/video/x2uemm4, which accentuates the heights of the bars over anything else.
And this:
I found this video and others, all of which are sonifying the y-axis rather than the x-axis: https://www.highcharts.com/demo/highcharts/plotline-context I’m starting to think that the shape is based more on the y-axis (height) of the bars rather than the position on the x-axis. Maybe we can incorporate both?
Also, I made this, which is pretty cool. I didn't know desmos could sonify functions.
Patch from 11/19 pairing:
Updated patch from collaboration with @matthew-blackman. Please bring the sound files from CAV over too (not in the patch):
During the 2/20 sound design @emily-phet and @Ashton-Morris, we discussed ways to simplify the histogram sonification so that it is easier to attribute a sequence of sounds to the shape of a histogram. We agreed to try the following specifications for the sonification of the histogram:
@matthew-blackman and I collaborated on the implementation, and incorporated feedback from @emily-phet @Ashton-Morris @catherinecarter and @ariel-phet.
The updated sonification plays a single tone for each column with the pitch mapping to the height of the column. The delay between tones is proportional to the width of the bins. We tested under a variety of circumstances and it seems like it is a good direction. Several things could be adjusted such as the exact sound file, the pitch mapping and the duration mapping. We recommend not having "gaps" where there is no histogram data, in favor of a more continuous sequence of tones.
I have added some mockup videos as requested to our folder: Mockup Videos
Also for out records we will be using the sound file pdlCannonLandToneV2.mp3 currently in the histogram.
The histogram is sounding amazing! We incorporated the recommendations from @emily-phet and @Ashton-Morris and adjusted the pitch range/bin timing so that it has a good balance of playful and meaningful. Fantastic work all! Let's close this issue and continue fine-tuning this feature in https://github.com/phetsims/projectile-data-lab/issues/169.
Reopening to bring up the sonification of a histogram such as this one:
With regards to @samreid's comment above:
We recommend not having "gaps" where there is no histogram data, in favor of a more continuous sequence of tones.
I was expecting to hear the gap between the two pieces of the overall data set. I'm wondering if there would be value to opening the discussion for hearing the gap when there are more than 2 bins (or so) empty to reinforce a larger gap in the data set?
I'm glad you brought it up @catherinecarter! I think the gaps could work if we use a visual bar that sweeps across the data, and not just highlighting the columns. I wouldn't want too long of a pause without associated visuals in cases like this:
I'll also bring this up in the sound design meeting tomorrow. I think the gaps would help convey the correct proportions of a skew, so I see value here. But I wonder if it would sound strange or unpleasant if there were significant pauses between the sounds. What do you think?
I agree that a significant pause would sound like something is broken. So a short pause would suffice, I suspect, although I'd have to hear it to know for sure. I don't think the size of the gap would need to drive the pause length, so maybe a short pause – long enough to know there's a gap but short enough so it doesn't feel like something is broken – maybe 0.5 seconds or shorter?
I wonder if the current highlight of the bars in the histogram could just be a highlight on the x-axis (the length of the whole gap) at the same cadence as the current sound for adjacent bars would work to achieve the visual pause you mentioned above.
I think it sounds particularly satisfying when you hear the same data from different bin widths.
The remaining work here is being addressed in #174 and #179. Closing.
The sound design team agrees that sonifying the histogram would be a meaningful feature to improve the sim's pedagogy and accessibility. Discussing the histogram learning goals with @catherinecarter, we reached the following specifications that this feature should support: