�ݺ�ߣ

Editor's Notes

• #3: I’m NSA from Blink. Blink is a UX research and design firm located here in downtown seattle. Today I’m here to talk to you about Siri.
• #4: (30 sec)
• #5: INTRO (5 minutes)Most of you probably have seen this commercial countless times. It’s one of many released by Apple featuring celebrities, this one ZooeyDeschanel, using a service called Siri. A year ago, Apple released Siri into the wild after acquiring it from another company. Siri was presented to us as a potentially amazing new way to interact with mobile devices. In these commercials, we see people using their phones to do ordinary things, in an extraordinary way. Through words instead of taps. In these commericals, Siri seemed like a technology that would seamlessly integrate into our daily lives. That it would be able to look up the information we needed, remind us of important events, remember things about us– just like a personal assistant. -Siri made it seem easy to speak commands to your iphone. It seemed as though Siri immediately understood these commands and instantly brought up information that previous would have taken several seconds and brain cycles of looking and tapping. -Celebrities and regular people alike were shown using Siri effortlessly, as if we’ve all been using speech recognition for our entire lives. It seemed like an extension of typing or touching on a screenIn short, everyone had high hopes, and Apple’s advertising with celebrities like ZooeyDeschanel certainly provided fuel to the fire.
• #6: Now the interesting thing is, Apple didn’t develop this technology from scratch, nor were they the firstThere are two main technologies behind siri – speech recognition and text to speech input. And apple was getting themselves into a landmine that has been explored for half a century now. Slow progression – “like watching a child grow up”1950s: audrey – recognized digits spoken by one person1960s: ibm shoebox – recognized 16 words in englishText to speech1970s: DARPA (defense advanced research projects agency) research – CMU harpy – 1000 words (3 year old child); bell labs technology to recognize multiple voice
• #7: *** This is what Siri would have sounded like in the 1960s. (older VR stuff here)
• #8: Obviously not as elegant as what we’re used to today… but it’s a start
• #9: Starting in the 80s, SR became more commercially viable due to the development of a new algorithmhidden markov model = looking for patterns in the individual utterances that form a word in order to guess what that word might be -> several thousand words1980s: kurzweilvoicemed and voice – dictation software for medical and consumer industry; dragon systems (dragon dictate); had to speak very-slowly
• #10: This is where we were in the 80s
• #11: Advancements in accuracy started to slow around the 90s but improvements in HW pushed SR even further1990s: faster processors = better SR. could speak more fluidly; business apps; BellSouth voice portalEarly 2000s: VR built into the OS level (vista, os x); accuracy topped at 80%; starting to go to the cloud- google voice searchLate 2000s: in the cloudBut that’s not the only reason apple thought siri would be betterThey thought they could succeed where others had failed– specifically, in making SR relevant, useful, and comfortable to use-examples: kinect, GPS (awkward, inaccurate, frustrating errors) Siri relied on apple’s own brand of magic, their skill at making technology beautiful, simple, intelligent and seamless, -use without thinking about it-multi modal-contextualAnd so, when apple launched siri – they didn’t mention that it was bought from a 3rd party. They didn’t mention it was in beta. -they treated it as the crowning glory of the iPhone 4S. As a huge differentiator from previous models, and a reason for consumers to care. As a game changer.
• #12: Butwhat do everyday users think? This was a fasicnating topic for me. As a user experience researcher, I have spent over a decade understanding how users fit tech into everyday livesStart in speech recognitiion and TTS at att labs researchNow at blink When siri released the research team wanted to take a deep look at itAt Blink, we wanted to understand the adoption rate and patterns around Siri. We conducted a month of user research with 12 brand new iPhone 4S ownersThrough regular interviews and daily diaries we captured these new users’ thoughts and use of siri
• #13: Deeply emotional responseNot just trying, thinking it’s ok and getting on with lifeFrom this we learned -when siri worked well-what roadblocks stood in the way-how behaviors changed over time-how impressions of apple changed-how the first month really does have a long-term impact on UX Most important: what we as technologists and innovators must do
• #14: Siriwas one of the top reasons participants bought iPhone 4SCommercials, friends, SR apps = see how it comparedParticipants had some pretty big preconceived notions of Siri based on commercials and friends’ experiences. Would make text input easierWould reduce RSI Would make iphone personalWould make iphone fun
• #15: Zooeydeschanel makes it look like a fluid, seamless conversation with a friend or personal assistant. Interesting initial assumptions about siri – using siri gracefully-like talking to a person; refer to as “she”-can learn and remember things-has a personality-understands me-helpful and useful-understand most commands and accurately dictate lengthy texts and emailsNot just barking commands at a machine
• #16: During first few weeks of use, ppts were optimistic about what to expect and how to use Siri. participants explored broadly to try to understand it better Tried asking simple and complex questionsLook up info from weather to movie times to store locationsDictate short text and lengthy emailsSpeak in a variety of tones, speeds, level of enunciation
• #17: Also tried asking it for fun things to see how much of a personality it hadHiding a bodyWhat’s the meaning of lifeWhat do you look like Not important, but signs that Siri is more than just a command engine. That it’s trying to go further Users liked the idea that there were little surprises, like rewards for using it more
• #18: (2 min) After using siri more, users became frustrated, felt their hopes and dreams for siri were shattered:-Notas capable/robust as expected- it seems like Siri doesn’t always understand or support my needs-can handle simple, short commands -dictation quality is poor, unreliable-range of capabilities is unclear – what else can I do with it? had to experiment to find range of capabilities but users weren't interested in that anymore
• #19: Siri can’t open or pass input to third-party apps Especially for users who have a preference (urbanspoon instead of yelp)never imported music onto iphone in first monthParticipants wanted to control what app or source siri used for commands rather than sticking with iphone native sources-pandora-fandango-gps
• #20: Siri doesn’t handle schedule-based requests well-ferry schedules-mariners games-movie timeswaste time speaking this command, only to have it not supportedstill have to go somewhere else - usually to google or an app, and dig around for the infosiri was supposed to deliver this info in a flash
• #21: Participants LOVED using Siri when driving, especially for driving directionsgreat hands-free way to look up a location and get directions from current place. But, still had to use hands to review and use the directions bc no TTSThis seemed like a major oversight – siri can access native apps and send inputWhy not use siri’s voice to present output? Just like a real gps?
• #22: Want to just speak, not think about it, not change their behaviorWanted it to be reliable. effortlessDidn’t know why it didn’t work-do I need to speak slower? Louder?-am I asking for things it can’t do?-am I phrasing it wrong? Users felt like they were to blame. That it was completely out of control.
• #23: To do that, needed acceptance:-Siriis accurate only half of the time-misunderstandings and misspellings are common; Only good for casual texting and email (no punct, accuracy)-needs short, broad one-off commands (weather, directions, alarms)-only works with native apps-aside from driving scenarios, is not faster than tapping on screen
• #24: MOST COMMON: Hands-free use-driving--dirty/occupied hands--shopping-safe and legal to use the iPhone while driving. (attempt at justifying use)-reduces the effort needed to complete some tasks (RSI, older users)
• #25: Participants formed habits around using siri:-IN VERY LIMITED ENVIRONMENTS-- not in places or situations where they could be embarrassed-- not in noisy areas-not in scenarios where misunderstanding is embarassing (siricalling ex boyfriend rather than a friend with similar name)Despite frustrations,users wanted to keep using siri to “get their money’s worth” out of the purchase-will continue to use occasionally -Would miss if gone-helps reduce RSI, makes some tasks a little less tedious-Hope that accuracy will improve over time (experience, learning)
• #26: In the months since we completed this research, new iOS and siri updateSiri launch – lots of tweet mentions about expanded featuresCan open appsRead turn by turn driving directionsGives football scoresShows movie trailersLook up schedulesCan post dictated status updates to twitter and facebook“siri is finally worth using now”
• #27: 6 months laterBut let’s go back to the every day users… what do they think about Siri now? Do these updates matter?Brief phone interviewsExperience was so…-poor-unreliable-made them feel like they were doing something wrong… that they didn’t want to use siri againUpdates to siri didn’t matter – prove it to me first
• #28: What have we learned? IsSiri dead? The limitations and frustrations of Siri should not be taken as a sign that this technology will never work. It shows that there is something promising in the air. We’ve come a long way in the last 60 years.Technology IS at a point where it can help us. Siri is an amazing example of that. Speech recognition isn’t just being studied in academia. It’s not just being used in DARPA projects.It’s here, for consumers, for professionals. The key is : it can’t just be novel – can’t just be frustrating. Has to be a reward for going thru the growing painsWe have to get to a point where users are comfortable with talking to a computer. -Where it feels like talking to a person. -Where, despite miscommunication, you still get something out of the conversation. -example: where results are contextually relevant. Where it seems like the thing youre talking to has learned about you and is tailoring information for youI ask everybody here to think about this. What can we do to make using Siri, or using any technology, like having a conversation with an old friend? How can we make technology more fluid, more seamless, and more rewarding for users? What can we offer users to help them through these growing pains of innovation? What will make it all worthwhile for them?