Sentiment detection is fast becoming a popular method of deciphering Tweets, news stories and other content on the web. TechCrunch50 startup Affective Interfaces, which launched at the Business Applications session today, has built emotion sensing technology that detects a web user’s facial expressions via a webcam.
Affective Interfaces’s SaaS uses a webcam to monitor facial expressions such as smiles, frowns and other emotions. The software analyzes these expressions and how they related to different parts of a site. So for example, Affective Interfaces will compare the emotions of a user while they are on CNN.com vs. Gawker.com. Or the software could be aptly used to determine emotions on an e-commerce site, like Zappos.com, to see what emotions affect engagement and purchasing decisions. Affective Interfaces gathers customer content through video ads, media etc., then collect and process the data for clients and create a report that deciphers the emotions of users.
The technology seems similar to having an online focus group test out a product, except the software actually breaks down the emotional data. Affective Interfaces provides a full report and charts with each session that drills down analytics from product engagement. The startup says it is releasing APIs for people to channel their data into the software, with ideal uses of the software for online media metrics, entertainment, webpage development, gaming sites and e-commerce.
Expert Panel Q&A (paraphrased)
The experts: Satish Dharmaraj, Lior Zorea, Bradley Horowitz, Tim O’Reilly, Kevin Rose
SD: I didn’t get it. Felt more like a PhD thesis than a business.
TO: I had the opposite reaction. I think there a couple of businesses here.
KR: I thought it was awesome. For focus groups and market research, this is going to be huge.
A: We see potential for an iPod detector etc. We see a lot of potential for partereing for market research, ads. etc.
Jason: You do that through mechanical turk?
A: yes, but you can also see your sessions.
BH: We’ll soon be able to record every moment of our life. If you could measure heart rate, that would be easier. Have you thought about other modalities?
A: We are thinking about other biometrics and modes. But the nice thing about AI is that it only requires a web cam.
Michael Arrington: The videos are really compelling.
A: We’ll be able to do statistical data sets on particular segments of the market. We can correlate frame by frame what the person is feeling and how strongly.
Video:
Outside Coverage:
TC50: Affective Interfaces detects whether your ad makes people happy VentureBeat.
TechCrunch50: Business Apps on Parade (and Emotion Sensing!) AppScout.
Affective Interfaces’ Webcam Software Reads Emotions Gizmodo.
more privacy killer shit.
well if you watched the presentation we make a statement about ethics and protecting privacy right up front. we’e also paying for the data. if you go to our data collection site at affectmetrics.com you can read the EULA: informed consent. wed like to help folks with this technology, thus the statement and offer of our API.
We take privacy very seriously. If there are any concerns regarding our policy or use of emotion data in general we are interested in your feedback. Thanks,
Jai Haissman
Founder // CEO
Affective Interfaces, Inc.
Jai@AffectiveInterfaces.com
I am really happy to finally see a startup that understands how important human emotions are. All entrepreneurs should know that great products appeal to our feelings, technology is secondary.
How long before we see fine print in EULAs and privacy policies (well, anyone who actually reads a EULA or privacy policy) that say any webcam on your computer can be activated and the information gathered can be used by the website without further notification. . . Suddenly an external webcam that can be unplugged starts to sound better.
I was thinking exactly the same thing –
Of course, what makes this sort of thing crazy attractive is the fact that there are cameras built into all sorts of portable devices, and most people do not know exactly how to turn them off, even if they can be turned off. Besides, an offer of $5 off the next purchase is often all that’s required for people to sign up for practically anything – brilliant!
reasons to spy on people with their webcams have been around for some time, and protections are built in at the OS level. if there were ways around we would already have seen it already.
http://affectiv...om/Welcome.html – link is broken. This is the second link I’ve hit from you guys that’s broken today. Slow down and get it right.
Thanks Eric, in progress.
This would be a great application in Cars, security to detect potential danger. May be his nervousness got the better of his presentation, but he kinda contradicted himself by saying this would be moving away from focus groups, but then it relies on users allowing the application to capture their responses through webcam. Again, I don’t know how many ppl would be comfortable doing that, and without having sufficient data of the ppl being analysed, the results won’t be very valuable.
Hi Affected Interface,
We are working in the same area — complementary approach.
First let me say that you really did not get a very good panel discussion. This group of technologists is very ill informed regarding branding, branding purchase cycles, etc. — but I’m guessing you were thinking that while they rambled on.
Anyway, right now I would focus your efforts on attacking the established concept ad focus group testing. This will give you an immediate entry into some very large an powerful consumer products companies.
From this point you will be naturally lead by you partners in the direction which you spoke about at about the 11 minute mark re: prediction purchase intent for new products. That would be huge if you can be better than current (archaic) processes.
I *would not* talk about nor work on API and platforms for wide-spread consumer adoption. That is a total red herring right now. Very highly valued datasets and processes can be achieved with sample sizes groups. Sounds like you are thinking/saying things you want for VCs, or a misguided vision about getting big fast.
You can easily get to organic growth cash flow by pursuing the consumer brand advertising research path. You should be focusing on IP protection also to the extent you can.
Once you have some established partners and cash flow, THEN think about the API and consumer stuff. You will find yourself in a much, much better position. Your consumer brand partners, if you don’t already know this, are likely the world’s foremost respository about affected response and purchasing. They will be your VERY, VERY best partner because they will want you to succeed. Just be careful to read their CDA’s to ensure you retain IP ownership. A technologist VC is not going to be anywhere near as good.
Finally, re: your likely infastructure costs. Get to know CUDA real well, and get some Telsa teraflops from nVidia. Or, more likely in the future, Grand Central and Open CL with nVidia or other cards. You shouldn’t need but a few tens of thousands for this, but you’ve got to starting thinking about parallel (CUDA) approaches for the longer term.
Best of luck. If you follow this path, then I’d bet for sure your looking at $50 million minimum as an acquisition by a company like TNS, but likely more. You might also become very, very interesting for a company like Amazon if you can get that purchase intent thing working well. Note: I do not think you will find much success with the Mechanical Turk approach for good research samples. Your big consumer product partners will take care of all that, so you can focus on your tech.
Wish I was a VC so I could invest. But, for now I’ve got to get back to my code. Make sure you get that message out regarding the importance of emotional response. Thanks.
Sounds like a great advice, I wish there were more high quality comments like this on TC.
Steve, –Thanks for the thoughtful remarks and suggestions. Would be interested to hear more about what you are working on.
The advantage of the API is seeing where best use lies and supporting deployment. We see great applications in attuned tutoring systems and therapeutic tools a well as business applications, but don’t have the resources to execute on all fronts.
So instead, we are building an architecture that will accommodate developers to use our insight for their own good work, creating a virtuous circle and making the data accessible to many rather than just a few. We believe in helping companies make better product and also believe in helping people become more emotionally aware to being manipulated, thus the desire to support emotional development applications (see below).
Although you are correct we can perform metrics on representative populations, there are many things of greatest value not addressed in the short presentation time frame: the long term value in a platform play, and other opportunities. We’d prefer to offer this utility ASAP while also gathering massive data sets which create value on multiple fronts, many of which I didn’t mention.
Here’s a partial list of ways developer partners can work with our API:
Attuned Machines
Computer responds to interest / frustration
Netflix Affect Based Recommendation Engine
Hulu Audience Metrics
iPod Attuned DJ
Photo Emotion Analytics
Mood Home Lighting
Nielsen Set Top Boxes
Fun:
Emotion Engine for PS3, Sims
Video Chat emotion sensor
Poker training tool
Attuned Tutoring systems:
interest / frustration tracker in dynamic learning systems
Emotion Development Tools:
Improve EQ
Anxiety / Depression Feedback
and Support Tools
Drowsiness Detector:
Auto
Transportation
Safety Professions
The product map is too broad for us to address just ourselves, so we’d prefer to partner for the greater deployment and utility. As the relevance of emotion for best design becomes widely recognized we will be positioned to help many. Glad you saw the value right away.
sorry, but there’s no way this works. we just don’t have the ability to determine affect on-the-fly over webcam. unless they have some major ML breakthroughs, it’s just so difficult right now. our technology sucks. we have FACS but it’s just not good enough.
I’m sorry to hear about your technology! And thank you, our technology is quite a breakthrough. Our accuracies are established, with high human expert inter-rater reliability.
Tim O’Reilly is less interested in detecting your sentiments than he is in detecting your usage of the term “Web 2.0.” So he can sue you. For using a meaningless marketing buzzword he claims to have coined but didn’t.
Yeah I wish I had it now to detect my expression of utter confusion. As in WTF is this good for really?
Affective Interfaces will be the second product out of the gate on this, since FaceReader is already being marketed and sold by Noldus Information Technology. It was released in 2007.
Affective Interfaces is extensible, tracks more axes, has a platform with API, and a focused solution for market research. Our sensors are proven quite accurate and robust. We look forward to benchmarking against Noldus.
Noldus has multiple software applications. All we do is affect sensing solutions!!
This is very specialized specific technology that will thrive in its niche but its not gonna change how 99% of research gets done today in that we still need to talk to actual humans to get them to tell us the story of why experiences with brands/prodcuts/services suck or rock.
Its great for advertising folks, provided it tells them how to make a better ad. If this just collects emotional cues but never figures out how to give the researcher a means to probe, rewind and play back those emotions, let alone asking a user hey why did you blink right there at that moment in the experience of x that is gonna be a fun ride for the researcher.
Cool core tech but less of a biz. KR saying this thing is gonna be huge i think is a reaction to the coolness of the tech itself and less of its real applicable use.
I doubt we’re ever gonna see a time where Zappos asks us to enable our webcams freely on own accord to then buy a shoe on their site other than if we were paid to do so via usability test, if so then yes, again to that specific niche it could be handy but then it doesnt tell you everything. Its like eyetracking trends out there, cool tech but alone doesnt tell you everything.
Maybe this tech would rock in the UK where there is a camera every 10 feet or so they tell me, even so, there will be dozens of players in this space.
To win they need to break out the tech and provide people a platform to dev from it. Open up the possibilities and let the masses and niche devs to show you whats capable with your tech and ride that way till something leaps out.
Hey Dan, thanks for the thoughts!
The accuracies and robustness of the system, patents, and our methods of analytics and deployment are significant, protected, and the insights are highly granular and market relevant. Emotion is compelling for purchasing intent and execution, and our methods of objectively capturing that extensibly over the network are unprecedented. Preconscious emotion is powerful but has not been reliably measurable across large samples until now. The unmet need for quality objective insight from a representative sample without geographic restriction into what makes products compelling (or not) has been sought by product development and creative teams and is now available through AI.
We believe the future of behavioral marketing is in informed consent, protecting privacy, and incentivize for data.
Our beta partners are already asking their customers to enable their webcams for just that reason! And we are successfully collecting data through these means.
Naturally we do not replace other useful research on what markets think of product or brand. We don’t tell you everything, we tell you what is fundamentally interesting and presently unavailable through scalable means. What we provide is the missing piece of what is going on in the preconscious and unconscious mind of the consumer, the things that determine behavior but may not be evident to the consumer or researcher presently. Insight on what makes product or messaging truly compelling will enable our partners to make truly amazing products.
Thanks for the interest in using the tech in other applications. We see a huge amount of potential here given the centrality of emotion for economic and quality of life considerations. Top support deployment and realizing the utility beyond what we can serve we’ve built an API in beta for those who would like to use our platform to make what cool applications they can imagine. We’re excited to see what’s possible when machines can sense how you feel and respond appropriately.
This is the first TC50 company that has some seriously cool technology behind it.
@steve above has already told you all you need to know about what to do business-wise.
Good luck!
reasons to spy on people with their webcams have been around for some time, and protections are built in at the OS level. if there were ways around we would already have seen it already.
Nice article………