Dear corporations: We've already ceded our privacy. Now implement the technology to serve us right
I honestly believe data can save the world if we get out of its way. This getting out of the way of data goes by another name: transparency. While I deeply respect the people who still fight the battle for our privacy, privacy and the Fourth Amendment are pretty much dead. The “man” will always be watching and will always find a way to justify a search or seizure. We, the people, might as well enjoy some of the benefits of our burgeoning, “1984”-ish surveillance state.
That said, like anyone, I’m often creeped out by items the companies send me that indicate they know a little too much about me. However, I’m often equally surprised and annoyed when they don’t. The Dead Kennedys said it best: Give me convenience or give me death.
Last week, I wrote that our lack of imagination in applying machine learning is stopping us from solving significant, practical problems. This week, I give you 10 real-world examples of hassles that could easily be avoided if companies that are already tracking our data simply connected a few dots to our benefit.
Solution No. 1: Simple rules engine for flight delays and rerouting
U.S. Airways/American Airlines lacks the ability to rebook your travel itinerary if a delay causes you to miss a connection. Delta not only detects this, but it will automatically rebook you on a terrible flight with two extra connections that will get you to your destination a day later than originally booked. Sure, you still have to call Delta to get a nonhorrible flight that might conceivably be worth close to what you paid, but at least it knows you won’t make your connection. This isn’t even machine learning. It’s messaging with a simple rules engine combined with search.
Solution No. 2: Sentiment analysis on social media
Marriott can’t tell whether what I say about it on Twitter is good or bad, which can be pretty hilarious, as the following exchange shows:
The housekeeper, whom I had tipped $20 for being awesome despite my kitchen messes, was presumably castigated because of what at first I thought was bad sentiment analysis that couldn’t process “unusually good.” But Marriott said it was a person, not a bot named “Janice,” who read the tweet too quickly.
Why don’t more firms use at least a phase of sentiment analysis in their social network monitoring? Sure, they will need some human error-checking because my friends and I will most certainly mess with any bot to expose its hilarity (#geekfun), but natural-language processing (NLP) and sentiment analysis have come a long way. They could at least color code the tweets for the operator. Heck, TweetDeck should do this.
Solution No. 3: Simple Web service between airlines and car rental companies
Hertz partners with airlines, but the airlines do not automatically pass my flight info to Hertz. Sure, they should ask my permission, but I can’t imagine why I would hide that my flight changed from the rental car company. This is either a Web service or a simple pub/sub, possibly some standards-making. But keep the semantic Web people away because they never put out actual software, only more academic blather.
Solution No. 4: Syncing up services so that you never run out of gas
Waze and Automatic don’t talk. Automatic presumably knows you’re running low on gas. Waze knows where the nearest gas stations are and whether you have enough gas to get there. Google knows if the gas stations are open. Of course, Waze is another (recently acquired) Google product that has bad search — like the forever ironic Google Docs/Drive/what have you suite — so it might not be able to find a gas station.
The red empty fuel light nagging me isn’t enough, and I usually have that covered up with my phone anyhow. This is simple messaging, search, and an event.
Solution No. 5: Simple machine learning to discern dietary preferences
The Caviar delivery service thinks I want a cheese pizza. It basically ran a promotion because I’d ordered delivery in New York. However, I ordered vegan. I searched on vegan. I also ordered delivery in San Francisco. It should probably know I was in New York for only a few days. It should also know that a cheese pizza wouldn’t interest me. This might be a recommendation engine (other people searching for vegan food also searched for X or never ordered Y) or a machine learning-style grouping problem (vegans don’t order dairy cheese pizzas).
Solution No. 6: Medical history screening app
I’m stuck in an airport lounge, but I’d rather be here than at my doctor’s office. For one, why can’t the doctor’s office message me and let me know not to come until 30 minutes later, given that my doctor is running behind? That is simple trend analysis.
Second, I already know that the only possible treatment for my Achilles’ tendons given what hasn’t worked thus far is a platelet-rich plasma injection. Why can’t they have me download an app where I answer the same stupid questions over and over and over and apply a rules engine, rather than making me come in to the office? We could agree that I’ve been “seen” virtually, not waste two hours of our time, and they could charge me the same.
With NLP, simple forms, and a rules engine, you could replace nearly everything about my initial evaluation for most anything that has ever ailed me. In fact, a simple Google forms tree would probably eliminate half the “I have a virus” visits. Get your blood pressure done at CVS or some place with a shorter wait, and it could be transmitted. With further analysis, you could even tune the treatment for all patients based on outcomes. Of course, that would be efficient and we’re talking about the American health care “system.”
Solution No. 7: Preference matching for the “sharing economy”
Uber should do more driver/rider matching. In denser areas where there are a lot of drivers and a lot of riders, matching various characteristics is important. For instance, a driver who smokes will likely do better with passengers who smoke when s/he “doesn’t smoke in the car.” I mean, if Uber is going to post and delete creepy stalker posts, it could at least rip off technology from dating services. It could even figure out how long I’m willing to wait for various matching traits (such as nonsmokers). This is either a basic recommendation engine or simply search.
Of course, a fancy graph search could help me see who my friends of friends liked and didn’t like as a driver. But then we’d have to stop pretending that all Uber drivers drive carefully, don’t smoke, and clean their cars. At least you know that’s what you’re getting in a taxi, plus some texting while driving.
Solution No. 8: Sentiment analysis to scoop earnings reports
I recently demoed a distributed Monte Carlo simulation. I also recently went to Elasticon where the USGS explained how it uses Twitter to detect earthquakes. In thinking about this I realized you could probably use sentiment analysis and social graph searches to predict corporate earnings before they are reported.
In essence, are the people at the top of the company happy, are their spouses happy, are their assistants happy, their children? How does their expressed social media level of happiness compare to past quarters? My money says the CEO’s profile may be tightly controlled PR, but other people closely associated to the C-suite might reveal not corporate earnings but how happy they and the people close to the top are. I bet you could trade on this right before a really bright earnings report or a really bad one. I thought I was a total genius until a Google search revealed that people are already doing this.
Solution No. 9: Geolocation app for pre-emptive pizza prompting
I try to eat well when I travel. I stay in hotels with kitchens. I bring a chef’s knife. However, I am often so tired after a day of working in the data mines that I flop down on the bed and order a pizza. Presumably the pizza chain ordering app could detect that I’m not at home, that I didn’t go anywhere but the hotel and an office building — not to Whole Foods or Trader Joe’s — and ask if I want to order the same pizza that I ordered last time to my hotel room. This is geolocation data analysis and a push notification.
Solution No. 10: Grocery shopping history data that serve customer interests
When I got home, Harris Teeter — my local grocery chain that allows you to pick up stuff — should have offered me a basket of the same organic produce that I bought last time. It should have noticed I’ll occasionally buy the poison-coated stuff if it’s out of the organic. This is buyer history, recommendations, and geolocation, or maybe the same trend analysis it does for its own inventory management. I mean, for years stores have gotten us to sign away our privacy for a “discount,” so why don’t they have all the data they need to tell us what we were about to buy anyhow?
Look, we already traded privacy so that we can be endlessly connected to the Internet, get directions, receive “discounts.” Isn’t it time the overlords started putting that data to use for our convenience? The NSA surely knows what I’m having for dinner before I do. Why doesn’t my local grocery store or restaurant?
What are you corporations even collecting this data for if you don’t have the imagination to use it? You know where I am. You know what I’m doing. Pull my strings and I’ll go far!
