I’m almost done with the alpha release of my project, a web proxy that helps people with disabilities use the web using head tracking:
The problem is that I need to be able to collect some fairly personal information when someone submits a bug. This data is important to know, because it will help me train the machine better. I could automate this by asking people to submit a screenshot (including their face) and just pass it through a machine learning model for inference.
There are a few issues with this, some potentially legal:
- I would have to store photos of people somewhere. My initial target audience are people with the most severe disabilities (ex, those who have suffered a stroke or tetraplegia) so on top of dealing with sensitive information I have to be mindful of 508 regulations etc
- I would have to develop a model to infer this data, and I’m overloaded as it is
- I have no income, so I wouldn’t have the means of storing this data anyways (I suppose I could use Firebase storage or AWS)
- The number of people willing to submit bugs will be lower, thus everyone’s overall experience is diminished
Therefore, instead of dealing with all that I’ll just ask users to answer a questionnaire. My method is by creating a bunch of checkboxes with Emojis.
TL:DR; Is this form politically correct and if not and how can I improve it?
In order to improve the accuracy of our computer vision models, please answer as many of the following as honestly as you can. This information is not shared with anyone nor is it collected long term (not even in aggregate), and is deleted once the ticket is resolved.
What environment are you in:
What was the expected outcome? ___________
What actually happened? ___________
Your questions don’t seem offensive to me, but if you explain up front why you are even asking certain things you will likely get a beter response.
Why does my skin color even matter?
On your earlier comments, not sure why you need people’s photos or how all of your software works. Might help us if you explain that too.
Actually, the whole point of this post was to specifically not require a photo. Because there currently aren’t many free, no-app, no-setup options for people with severe disabilities to use the web (in fact, I don’t know of any others) I expect to have a lot of users and I definitely don’t want to waste hours going through bug reports of people trolling me with silly photos and for the other reasons I listed.
As for skin tone, it matters quite a bit actually! Imagine a green screen, where instead of it being green it’s the color of your skin. As you can imagine, it would be fairly difficult to pick out:
The problem is that the background environment doesn’t actually have to be the same color of your skin - things like ambient lighting may effect the perceived color so that the net result is that your skin and background appear to be the same color. Here’s a visual of the same exact White House in white, yellow, and pink:
These are very extreme examples, but help illustrate why it’s important.
That’s why I also ask for environment lighting, and whether the person is indoors or outdoors. If I get a person saying that they have a darker complexion and they are outside in daylight then that’s a major problem because that’s an ideal situation (from a computer vision standpoint). If I get another person saying the same thing but that they are indoors with poor lighting, then I might set that bug report as a lower priority.
I actually don’t expect to get very many bug reports but because this project serves people with severe disabilities, it’s important for me to collect as much info as is necessary (given that everything happens through the webcam)
Thanks for the questions (it helped me think this through a bit more)!!
You just answered your own question.
So instead of being “politically correct”, try telling people what you need adn why you need it and how they can benefit from helping you out and you will be down a good path!
I’m really not sure how to respond to this and I’m not sure what you mean by “See?”…I mean, I was answering your question lol
The reason for the brevity is so that the whole form can fit on the same screen. However this gave me some ideas to add tooltips next to the questions so that you can see why the info is collected. Maybe this is better?
As this is largely an accessibility issue, I’d have a few queries over the current set-up.
Can the icons be tabbed to and selected by keyboard? They present a pretty small click target for anyone with limited motor skills. Also, they’re quite small and not all that easy to distinguish the details for those with poor vision. Remember, too, that tooltips by default only work on hover and again are useless to keyboarders or those with poor motor skills. Ensure any additional information is available to all.
Skin Tone: I can answer that.
Eyewear: Hmm … I’m wearing glasses, but they’re not dark glasses. Should I go for the first one? Maybe that’s only for sunglasses?
Facial hair: What if I’m a woman and / or have long, loose hair over part of my face?
What environment are you in?: This one has me thoroughly puzzled. I’m in a house, rather than an hotel or a block of flats, or an office block or whatever that may be. OK - not sure why that matters, but I can decide between those two. But what’s the third option? Outdoors? Or rural area? It’s not clear to me what information you’re trying to gather here.
Environment Lighting: I could take a good guess at that.
as @TechnoBear says emojis could be interpretted incorrectly as they are pretty small and a bit vague. I’m not sure of the accessibility side of it but you could have a click to open a modal with a text description of each with a picture so the facial hair for example could have actual stock images of people with different hair styles. The tool tip could be included as most modals have the option for a caption.
As for political correctness i don’t think it is politically incorrect to ask for this information as it is needed to make the product work correctly and make improvements. If you were selling toothbrushes and asking for skin colour then fair enough you have no need to ask that as it is irrelevant.
As said above just make it clear why you need each bit of information and you are covered. You might get the odd one who kicks up a fuss but as far as i can see you have treated everyone equally.
Thanks for the responses!
I don’t know why but I haven’t really considered mobile at all for this form even though I talk about mobile a lot, thanks for bringing this up! And you’re right about the tooltips. Maybe I can make the icons way bigger, and show only one question on the screen at a time, something like:
The dashed line is the
<abbr> tag which shows a tooltip (I’ll be more explicit).
Great point on the facial hair as well. Perhaps I can try something like: (except that it would match the selected skin tone).
I’m going to rethink this form a bit as I’m really stressing myself out over something that will (hopefully) rarely be used. Thanks for the responses everyone!
Your service should help people with disabilities. I think they will be happy to participate in the study. I recommend trying to communicate with clinics and negotiate direct testing. The survey will not give an accurate result.
This topic was automatically closed 91 days after the last reply. New replies are no longer allowed.