We also have a new chatbot that leverages the new system that you can test here9.
I would love to get your thoughts on how it works. Try putting in your real needs, wants, hearing loss level, goals, etc… and see if you get a reasonable response and Let me know!!
Search string: Hearing in background noise with iphone compatibility.
Returns the Sphere. So that means the weighting towards devices where iPhone is specified doesn’t exist or isn’t associated.
We haven’t favoured Sonova products for a while due to the legacy charging and v5 receiver cerusheild issues. I’m not sure how you’d imbue your engine with ‘the experience’ to know this stuff.
Also the Sony seems to ‘win’ every other category; how do you account for/remove the consumer self-selecting purchase bias from the situation. I bought a Mercedes, therefore Mercedes make the best cars: even though other cars might be faster, safer and more reliable…….
Returns the Sphere. So that means the weighting towards MFi devices where iPhone is specified doesn’t exist or isn’t associated.
I don’t know that I would necessarily weight toward unless someone specifically asked for it, but it is a bug that the system isn’t finding MFi even when you search for it directly. This is because MFi is nested under Bluetooth and I haven’t handled that yet. However, we do have the data in the system.
We haven’t favoured Sonova products for a while due to the legacy charging and v5 receiver cerusheild issues. I’m not sure how you’d imbue your engine with ‘the experience’ to know this stuff.
I don’t know that that is universal.. I know plenty of happy Sphere owners, and it tested well in our lab conditions. Regarding the collective experience of the forum users, that’s something I’m working on for the next iteration… you can accomplish this through a RAG query system.
how do you account for/remove the consumer self-selecting purchase bias from the situation
If you try the bot, it asks about your hearing loss level and budget, which determines whether you are shown for now. The plan is to introduce more questions so the bot is able to determine OTC candidacy more comprehensively. I didn’t want to put too many questions in front of the recommendations at the end though, while testing the flow. For the moment, you can simply add to your query that you’re looking for prescription hearing aids or include that you have severe or worse hearing loss to exclude the OTCs.
When I click on the link you shared, there is no bot? I guess I feel that when most people see “hearing aid” recommendation they are trying to find a hearing aid, not an device, and so wouldn’t think to write “prescription” in the box. Maybe the simplest thing would be to just say, “specify whether you are interested in prescription, OTC or both” in the little blurb.
Sphere is a honker and the charging case only does two charges.
This time I just flat out asked for what I wanted and didn’t get it:
I’m not disagreeing with you how it’s set up based on the parameters you’ve established.
However it’s like asking a question about the music you want played on a playlist without any legacy interpretation or experience of the situation.
So: frame the question, what’s the best ever single hit song? Put all the criteria in you like to establish this and see what falls out. Would it land the same as my choice? I doubt it, because you don’t arrive at the same situation with my experience and conclusions.
That’s even before you consider the actual effect of proper programming or self-selecting consumer bias (Mercedes argument).
I’m sorry, but: ‘Computer says X’ is a really bad prescription model.
I tried various descriptions of what I want in a hearing aid. But unless I listed prescription or accessories, I got Sony as my recommendation. Even when I listed comfort? I was unaware an earbud could be comfortable for long days. Also I don’t want an earbud, but would not have realized I needed to say do.
I guess I feel that when most people see “hearing aid” recommendation they are trying to find a hearing aid, not an OTC device, and so wouldn’t think to write “prescription” in the box. Maybe the simplest thing would be to just say, “specify whether you are interested in prescription, OTC or both”
Good idea… I’ll add that and just generally more about the caveats, like that we don’t have information about colors or sizes…
Sphere is a honker and the charging case only does two charges.
We don’t currently have sizing, but this could be added. We also don’t currently have structured data attached to the devices about rechargeable battery duration or charges in the power packs… but have plans to add, as well as more about what specific features the apps have.
This time I just flat out asked for what I wanted and didn’t get it:
It’s not a search engine, so I’m not really surprised. The main goal is to help newbies get oriented through the bot… this querying system will be used by the bot to serve up product recommendations after the bot gathers as much information as possible from the user.
So: frame the question, what’s the best ever single hit song? Put all the criteria in you like to establish this and see what falls out. Would it land the same as my choice? I doubt it, because you don’t arrive at the same situation with my experience and conclusions.
I don’t expect this to be a really useful tool for experienced users… maybe eventually after I build in more of the user feedback from experienced users.. but definitely not right now. As I mentioned above, the goal will be to help beginners get oriented … the market in the US is complex, and a lot of people are now trying to self prescribe with … so this tool is primarily a tool to help people select an appropriate OTC (or at least avoid the junk ones)… or to get them on the road to clinical treatment where of course an audiologist will become involved in the recommendation process.
Well this is good. With the prompt “Long battery life, excellent speech in noise, stable Bluetooth streaming, superior wind noise suppression” it came up with exactly the hearing aid I’m considering.
Can you use our audiograms as inputs? Rather than having to say my hearing loss, can you just slurp it in?
Not yet… but that is something we’d like to do… might be easy to say grab it from apple health. I’m not sure how many people store their there though. Otherwise it will be inputting it manually or trying to use computer vision to read a photo of your audiogram.
Long battery life, excellent speech in noise, stable Bluetooth streaming, superior wind noise suppression
Just a note that we don’t currently have rich data on “long” battery life though Steve’s ratings account for that and probably contributed to your match. Wind noise is also something we don’t measure, but we document which hearing aids have wind noise suppression, which of course is most of them these days. The “superior” part the system can’t help you with… We’re hoping to glean more of that information from the user feedback here on the forum over time.
Not yet… but that is something we’d like to do… might be easy to say grab it from apple health. I’m not sure how many people store their audiograms there though. Otherwise it will be inputting it manually or trying to use computer vision to read a photo of your
I was thinking more along the lines of pulling from a user’s forum account. Many of us have input the data to create our , feeding that into the engine would seem like a great first step.
That’s fair enough, however: knowing a little about the human condition and having experience of how people respond to the news that they are going to have to make a grudge purchase and overcome their denial: I think producing what looks like an ‘independent arbiter’ holds massive risks.
As soon as people get ‘a result’ (however uninformed it is) that will be their desired first solution: whatever they subsequently get told by an experienced professional. In this case the Sony non-custom : with zero knowledge of the impact of the loss type, canal shape/size, occlusion, directionality or longevity. Which just adds another layer of uninformed noise to the complexity of the fitting process with an undeserved level of authority.
Nobody coming in to my premises would get the Sony as first choice. Ethically, it’s messy to default to that recommendation as a function of incomplete knowledge or through consumer self-choice ‘satisfaction’ criteria. Or to put it another way, a customer is always going to exhibit bias towards their own horse: they literally have their own money invested in it. That’s the diametric opposite of a blind product test.
Hey @Um_bongo — just to clarify, this interface is purely for beta testing. It won’t be exposed to real users in this form (other than here in the forum and a few colleagues). Instead, it will serve as the backend for a chatbot that guides users through a decision tree before submitting their needs and desires to the product-matching engine.
The scenario you outlined won’t actually happen. Only users who meet clear criteria (U.S.-based, 18+, mild to moderate hearing loss, no red flags, interest in self-treatment, etc.) would ever see recommendations for products like Sony. Those users won’t be walking into your clinic unless they choose not to follow through with the bot’s recommendation. Or maybe after failing on their own and seeking professional guidance.
Any recommendation will be accompanied by clear disclaimers, especially about limitations in the availability or applicability of data, lab tests, etc. I’ve put a lot of effort into making sure that’s the case and will continue to refine both the data and the messaging (your help is welcome on getting the messaging right).
The goal is not to tell people “this is definitively the best product for you,” but to help beginners take the first step in exploring their options—especially in the OTC space, where there’s a lot of misinformation and low-quality products. Our lab testing is specifically tuned to mild-to-moderate , which makes us uniquely equipped to help people who are pursing OTC options without professional assistance.
I’d welcome any input on how to present this information more clearly and responsibly. This is just the beginning of the project, and I’m passionate about the potential to help people who are starting from ground zero. I fully understand your concerns and welcome the critique, but I’m committed to seeing this through. Being based in the UK, you may not have seen the scale of misleading marketing and scammy products we’re dealing with in the U.S.—it’s a real problem, and HearAdvisor is the only lab actively trying to bring clarity through independent scientific testing. I’m excited about making our data more accessible to the people who need it most
As we test more customs and in-ears, and begin testing more severe hearing losses, our data will have much more applicability for experienced users, but for the moment, we’re mostly focusing on the OTC market. Having said that, there are still millions of people in the medical market purchasing for mild to moderate hearing loss, and these are people we feel we can help as well.
Anyway, thanks for engaging on this, and again, I welcome your constructive feedback.
Auckland, New Zealand
and Let me know!!
Wales, UK
North West UK
San Diego, CA, USA