“We like proving that the stuff we think is awesome is worthwhile,” explains Charlie McKittrick, co-head of strategy at Mother, the ad agency. “My greatest fear is of being full of shit, or selling vapor product,” he tells me in Mother’s midtown office. “We wish we had a measure for success in experiential marketing.”
What kind of experiences are we talking about? Imagine meeting Little Marina, the fashion blogger from Varese, Italy who hyped Missoni’s 2011 collaboration with Target. Marina flew out to New York’s Fashion Week on Target’s dime, and turned out to be an enormous marionette. She handed out huge business cards listing her Twitter handle and phone number for text-messaging, and held real-time chats with everyone from Harper’s Bazaar editors to super-models to crowds of kids. She carried a free, open Wi-Fi hotspot. News outlets covered her globally–“2.8 billion media impressions to be precise,” as Mother’s publicity folks say, “which made her the biggest blogger Fashion Week had ever seen.”
So Marina had an impact. But was she worth the millions of dollars she cost Target?
Circle Media, the venture McKittrick pitched at SXSW in March, aims to measure brand engagement at events. Who comes? How do they respond? What do they say or share or photo-shoot? Do they change their opinions, how they shop, or what they buy?
Ideally, Circle Media wants to boil all these factors down into an index. Similar to what Nielsen’s produces for conventional advertising, this easy-to-read number, like IQ, could tell ad-makers and their clients what degree of cultural splash a particular event creates. Circle Media calls it “Event Performance Index” (EPI).
“We bring the measure to unmeasured media,” the company’s CEO, Mark Piening, explained there, where his company placed second in the Austin Startup Pitch Competition. “We want to help brands understand what audiences want,” he said.
The key questions, as McKittrick sees it, are: “Did I get the right people? Did they share what they saw? Did they hear the story I wanted to tell?”
What you see when you look at the Circle Media “Audience Intelligence” dashboard is a reflection of these questions. The metrics are called “Scorecards”: The first score, “Audience,” tracks ticketing data, as well as aliases gathered from attendees’ email addresses or Twitter handles, and their professional affiliations–e.g., news media, tech-developer, fashion figure–inferred from public Twitter self-descriptions. The second scorecard, “Sharing,” aggregates volumetrics on Twitter, Facebook, Instagram and other media-sharing platforms, and distills these into one metric of social buzz. Finally, the third, “Preference,” addresses the content, not just numbers: what people said about the event, and what they paid attention to.
This last metric relies on natural-language analysis software to classify statements like “Dude you gotta see this,” or “I hate Apple” into positive, negative, and neutral-brand statements. Joe McCann, the creative technology director, is also working on visual-image recognition software: algorithms to identify every time a company’s logo or product is photographed on Instagram. Training an algorithm like this is tricky for machine-learning, so he’s using the crowdsourcing program Amazon Mechanical Turk: Using Turk, he can pay strangers a few cents apiece to label photos that contain, for example, the image of a Microsoft computer, or a bottle of Coke. As his database of labeled photos accumulates, his algorithm becomes better and better at trolling Instagram or Facebook and tabulating every snapshot of a brand’s label or product taken at an event.
Yet another component of the data stream we used to call experience.
Such projects are attended by few people, but targeted at influential audiences. They depend on spontaneous social and media buzz–Twitter, YouTube, Instagram, news, and industry chatter (e.g., among the fashion or tech worlds)–to spread the message. This is “soft power” marketing–less intrusive, more friendly than conventional ads. But how well does it work?
The companies who hire new-generation advertisers want to know how much the sponsored events actually affect people’s thinking, not to mention sales. They don’t want hand-waving or pretty pictures. They want numbers.
“Mother Ventures exists not only as a means to diversify our business, but as a way to satiate our curiosity,” writes Andrew Dietchmann, founding partner of Mother New York, in an email. Dietchmann’s introduction to Mark Piening, then at BazaarVoice in Austin, led to the creation of Circle Media. Mother has invested money and resources into Circle Media, and has a minority stake in the company, but Circle is run independently by CEO Mark Piening. “In the case of Circle Media, we’re learning about big data as it relates to a core Mother capability–creating inspired experiences for brands. Circle will help us understand how to better conceive, sell, and optimize brand experiences by getting us closer to the audience.”
This isn’t just a fringe trend, either: The future of advertising looks to be not print or TV, but experience.
When you imagine advertising, what is your first association? Mad Men, maybe: The dated office culture of a sleazy business. Or, those lame clips of ex-golfers with arthritis who interrupt the YouTube clip you’re watching to prattle about a medicine that’s irrelevant to you. Or the list of corporate sponsors you fast-forward past at the beginning of your podcasts.
Even if you recall the ads later, your first impression of the brands is hostile: We don’t grant advertisers the right to interrupt our news or entertainment anymore, especially not to sell us stuff we don’t care about. We want ad people to know us better, and charm us with less predictable pickup lines, less intrusively. We’re forcing marketers to get more creative, and to pay more attention to us: to what we expose about ourselves by how we behave, online and off.
What this means is marketing is becoming more Big Brother, but also more fun. The lines between advertising, entertainment, social media, and journalism are getting blurrier, as the tools for publishing content, attracting buzz, and judging response, are becoming universal. A niche is opening up, meanwhile: How to measure the impact of “Experiential Marketing.”
Mother New York’s office looks like a mix of an art studio and a restaurant: The red British phone booth by the entry connects to Mother’s home office in London. There’s a taxidermied bear on the other side of the door. The lighting and mood of the place is cheerful, dress is hipster-casual. Plenty of Converse or Vans sneakers, tattoos, and colorful shorts, not unlike the vibe of a magazine’s office. The bar serves Stella Artois to visitors.
I’m sitting under a mosaic of photos of mothers of Mother employees, one in a swimsuit with a pregnant belly, perhaps containing a future Mother ad person. Another mother, I’m told, is a Playboy bunny. Mother employees’ business cards also have their mothers on the back.
Charlie McKittrick and Joe McCann continue to tell me how they imagine the future of ads: Advertising through events like Little Marina or Microtropolis, McKittrick points out, is much, much longer exposure than a minute-long TV spot, radio commercial, or magazine page.
“The stimulus has more data in it,” says McKittrick, whose Cognitive Science studies at Boston University with philosopher Daniel Dennet may partly explain his geek’s enthusiasm for hunting patterns in social datasets. “The brand has behavior and personality” expressed over time.
“Imagine Governor’s Ball,” says McCann, the developer who moved from Austin to Mother’s New York office in September to help build Circle Media. He’s talking about the New York City music festival. “People’s Instagrams are all muddy sneakers. You assume they’re in a bad mood–“
“Unless they’re hashtagged ‘#baller’,” McKittrick interjects from the other couch, here in Mother’s New York headquarters.
Each moment of an event–a music festival, say, or a fashion show–is packed with data on people, the two tell me. This data is readable from mobile devices: position; mood (from Twitter messages or Facebook posts); attention (what you’re Instagramming or sharing about). One can imagine a graph of an event’s social buzz, @mentions, # hashtags, Instagram shots, et cetera over time, in terms of any number of brands. Or, for that matter, bands. What are people saying about the Flaming Lips? Or MGMT? High-influence attendees could be tracked by their Twitter handles, to know whether the important people are paying attention.
That’s the future according to Circle Media, anyway.
Data Science is impacting media, as the technology for classifying brain-states and tracking people’s interests advances. Tweets, Facebook “likes,” eye-movements, brain-waves, and blood-flow in cortex: all inform scientists about what you’re feeling about what you’re reading right now. We’re here to report how this data is shaping the future of media, and how it may help long-form narrative survive. Armed with equal parts curiosity and skepticism, we’ll tell you what’s real and what’s over-hyped bunk.
July 15, 2013
Tweets during crises show how social news spreads online. A new tool called the Hashtag Crisis Dashboard boils Twitter hashtags down to line graphs, showing what people talked about when. Soon, it may help predict and avert disasters before they happen.
Knowing that lots of people tweeted “#frankenstorm” during Hurricane Sandy can’t stop the wind and rain next time. But many disasters, from bombings to political revolts, begin with Twitter buzz data scientists can detect. Real-time analytics of tags during an event could help emergency responders get there faster.
The Hashtag Crisis Analytics Dashboard, released this week, is designed to turn Twitter data into crowdsourced stories, readable by non-quants: Graphs of what the crowd is saying.
Tagged information tweeted during news events–Hurricane Sandy, the Boston bombings, the Egyptian Revolution, for example–could help stop the next tragedy before it begins, argues iRevolution blogger and “Crisis Mapper” Patrick Meier. Hashtag data on trending topics, analyzed and visualized, he says, can help journalists tell more accurate and compelling stories about how eyewitnesses respond to events, and speed up emergency response in the moment. During the Boston Marathon bombings, for example, the first tweets of “#bomb” and “#explo*” came within three minutes of the first bomb going off.
Meier, previously co-director of Harvard’s Program on Crisis Mapping & Early Warning, is now Director of Social Innovation at the Qatar Computing Research Institute (QCRI), which released the Dashboard. Through three TEDx talks and his popular blog, Meier has become a leader in applying tech to crisis prevention. And when he blogs, journalists pay attention: The New York Times posted a link to his blog this week when the Dashboard was announced. The tool offers great fodder for reporters, in an era when journalism relies ever more on data.
“What does Egypt’s Political Polarization Look Like On Twitter?” is the headline Foreign Policy magazine used to describe work by two of Meier’s colleagues at QCRI and one from Al Jazeera. By tracking the hashtag topics tweeted by people on both sides of the conflict–self-identified Islamists or secular opposition–the group could see which topics were neutral, and which were polarized liberal or conservative. #FF, for “follow Friday,” for example, was tweeted equally by both sides; #Coup was used more by the ousted Islamists, while the liberals preferred #revolution.
Using these metrics of “tag polarization,” the group was able to graph “Egyptian political polarization over time.” When superimposed over a timeline of recent Egyptian news–violent protests over military rule (April ’12), the constitutional crisis (Nov. ’12), and the recent Tamarod protests, on June 30– a story of waxing and waning tension emerges. If these surges of political polarization can be spotted, future casualties could be spared.
Twitter is most relevant in certain parts of the world. A recent study of geo-tagged tweets worldwide found that the top five locations with the most Twitter hashtags were Sao Paulo, London, Jakarta, Los Angeles, and New York. The study, whose data included 2 billion tweets and 27 million unique tags, also found that hashtags are “essentially a local phenomenon with long-tailed lifespans” and that those triggered by real-world events like disasters “spread faster than tags that originate purely within the Twitter network itself.” So, the hashtags that catch on are largely a reflection of local news, not Internet jokes.
The Analytics Dashboard’s debut showcases a demo based on Hurricane Sandy.
The interface uses three displays to track “hashtag footprints” of an event. In this case, the data includes 4,904,815 tweets, 1,194,885 of which contain at least one hashtag. They were gathered between October 27, 2012 and November 7, 2012, from Twitter users in New York.
The first display, called an “Annotated Timeline Chart,” shows Number of Tweets over time in orange and Number of Unique Hashtags in green. That is, sheer volume of buzz.
The next, “Hashtag Analysis” display, shows the Top K tags (10 is the default), as well as the Bottom K and Middle K–i.e. which 10 hashtags are being tweeted the most, the least, and in the middle of the range. There’s also a display of the Top K newest tags–what topics are popping up–and the “Top Increasing K,” the tags that are being tweeted more often now than before. These two measures are signals of new issues– like, say, a #subwaydelay.
Finally, the third display shows a plot of hashtags over time for each of however many hashtags you want to see. So, for Hurricane Sandy, you can see the rise and fall of buzz around #sandy, #hurricane, #noschool, etc. as overlaid colored lines. Different tags represent the perspectives of different groups–e.g., kids hoping for another day off from school; emergency responders; homeowners who have lost power. So the graph is a dynamic picture of the multi-layered social response to a public event. Juicy stuff for a data junkie.
Hashtag data tells a visual story about how crowds respond to news events online: Egypt’s political tensions; victims of disasters like Sandy; witnesses to crimes. Tweet patterns may even one day identify people who are violent, politically activist, or psychologically unstable before they blow up bombs, start protests, or assault political leaders.
But it’s urgent that writers don’t get so charmed by new data toys that we forget our job is to tell stories.
Take the Boston Marathon bombings. “Three dead, 264 wounded” is a statistic, not a story. The graph of Twitter responses displayed in a recent paper is likewise articulate data, helpful for emergency personnel. But at the end of the day, it describes tragedy as a line-graph.
To let people know what a crisis felt like, we still need stories like “Beyond the Finish Line,” Tim Rohan’s empathetic New York Times feature on what it was like for one 27-year-old to lose his legs and become a media symbol of the Boston bombing. Stories like this give readers the emotion behind the numbers, and show how a news event affected individuals. Journalists may use Hashtag Analytics’ crisis data to supplement our stories, but it can’t supplant them.
July 2, 2013
What if you could measure the moral slant of media?
You’d dip a digital thermometer into the script of, say, a Woody Allen movie, a broadcast of Rush Limbaugh or Jon Stewart, or a Michael Moore documentary, and see a readout of where it fits on the spectrum of values.
Morality metrics could judge spin automatically. Censorship apps could be made for parents to protect their kids, or viewers to screen for their own worldview. Policing apps could be made to detect ethical bias in programming. And, as John Voiklis is planning in his new position at Brown University, “ethical robots” may be designed, sensitive to the ethical tone of language: Bots programmed to take offense at insult, or to fight for good.
This moral “coloring” of words is what Voiklis studies for a living. On The Ripple Effect blog, the online outlet for the social impact research firm Harmony Institute, Voiklis reported this week the results of a study he’s been doing on the moral dimensions of network television. His coding system, coloring words in six dimensions, expands work on Moral Foundations Theory by psychologist John Haidt at the University of Virginia.
Voiklis’ moral dictionary now consists of around six thousand words–approximately 10% of an educated person’s vocabulary, Voiklis says–which three judges have spent hours labeling for moral coloring. Armed with this word database, Voiklis’ algorithm takes in a TV script and outputs what proportion of its words fall into each of the moral bins: this “moral fingerprint” can be compared to that of the average liberal, moderate, conservative or libertarian to determine a show’s or network’s political leaning.
UVA’s Haidt has found that moral language aligns along six axes: 1) Security (Kindness vs. Cruelty); 2) Justice (Fairness vs. Prejudice); 3) Community (In-group Loyalty vs. Betrayal); 4) Authority (Respect vs. Affront); 5) Purity (Innocence/ sanctity vs. Corruption); and 6) Autonomy (Freedom vs. Coercion).
“Morality space” can be mapped in these six dimensions–like Manhattan’s length and width, Voiklis says, plus four more–and the distance between two TV shows, movies, networks, or writer’s worldviews, can be visualized, almost like an address, in this space.
You can determine your own moral bias through the online survey YourMorals.com. The questions, each targeting a different dimension of the model, are like those on a personality test, or an online dating profile. You rate statements like “whether or not someone showed lack of respect for authority” on their “moral relevance” to you and others like “Respect for authority is something all children must learn” on how much you “agree,” both on a scale from 0 to 5. Each statement applies to a moral dimension, and overall the survey determines which values you prioritize in your moral outlook.
John Haidt believes the dimensions of morality are innate. But the relative importance of each dimension varies across cultures, and across political ideologies.
What his original Moral Foundations Theory study showed is the ethical coloring of political orientations. People who filled out the YourMorals questionnaire also gave their ideological sympathies–liberal, moderate, conservative, or libertarian. What Haight found is that, generally, liberals value “Security” and “Justice”, while conservatives’ moral systems are biased toward “Purity”, “Authority”, “Community.”
Voiklis’s new study for Harmony Institute has extended the political mappings onto TV content. His main finding: Contemporary television language differs most from conservative viewers’ values. This lends some quantitative support to Republican complaints that the media is “liberal-biased.” But on the other hand, conservatives were moral outliers: the pattern of their moral outlook differed not only from liberals, but from moderates and libertarians as well. This can be seen as quantitative support for the impression made by last weeks’ Supreme Court rulings on gay-marriage and immigration: that the conservative worldview has lost touch with modern times.
All this theory is very abstract, though. What does moral talk sound like? Let’s hear some hate speech for a taste. Here’s fashion designer John Galliano in his racist meltdown from February, 2011 in Paris, as quoted by Vanity Fair in its interview with him this month:
“I love Hitler. People like you would be dead. Your mothers and forefathers would be f***ing gassed and f***king dead today… You’re ugly.” To an Asian-Jewish couple at the same cafe, he said: “Dirty Jewish face, you should be dead” and “F***king Asian bastard, I’ll kill you.” As Galliano explained on June 13 in his first sober television interview, he was not only wasted, but deep in the throes of blackout alcoholism and addiction to benzos at the time. He’d been researching an anti-semitic figure whose epithets he says his drunken mind randomly free-associated to.
The morality-robots Voiklis is building might one day be able to show, from a record of all of Galliano’s public speech, that his drunken rants were a moral outlier–a color uncharacteristic of his personality and beliefs–or not.
Voiklis concedes that his approach to reducing linguistic meaning to numbers has its detractors. “In the humanities, there still persists this idea that ‘we are not measurable. If you try to measure me, you’re just being reductive.'” In fact, one of the judges Voiklis used to determine the ethical valence of 10,000 words from the dictionary is a humanist who is openly hostile to his approach. “But,” Voiklis adds, laughing, “she poured in two weeks of work in her free time for me. It’s always good to take in that skepticism from that circle of people, so you can show that you’re taking a multidimensional approach.”
Even if you’re open to a social scientific perspective on media, you can see how reducing the ethics of language to numbers runs into trouble.
Take, for instance, Kanye West, riffing about modern-day racism on his new album, Yeezus, in the song “New Slaves”:
“My mama was raised in the era when/ clean water was only served to the fairer skin/Doing clothes, you woulda thought I had help/But I wasn’t satisfied unless I picked the cotton myself/You see, it’s ‘broke nigga’ racism,/that’s that ‘Don’t touch anything in the store’/and this ‘rich-nigga’ racism/that’s that ‘Come in, please buy more./What you want? A Bentley? Fur coat? A diamond chain?/All you blacks want all the same things.”/Used to only be niggas, now everybody playing/Spending everything on Alexander Wang/New Slaves.”
These lyrics are packed with plays on words–sarcasm, cultural references and evocations typical of the way people speak. How could a machine understand the statement “clean water was only served to the fairer skin”, without knowing the historical context of Jim Crow? How would it get the bite of “they wasn’t satisfied unless I picked the cotton myself” without knowing about the cotton trade and slavery? How would Voiklis’ ethical robots deal with something like this?
We’ll be following Voiklis’ work moral robots at Brown, along with Harmony Institute’s latest in measuring the impact of movies, TV, art and news. Stay tuned for more as we update this story.
June 20, 2013
I’ve been a foreigner a lot. Japan was home for three years, France for six months in college. The rush of trying to stay on top of a fast-moving cultural wave is addictive. Channeling a foreign frequency–the alternative rhythms of a place that’s not your own–I’ve found transformational. You feel another culture changing you, your personality evolving, when you become a code-switcher. You develop new sets of symbols, new associations and priorities, new ways of seeing and expressing your self.
So I was fascinated to read about a new study, co-authored by Chinese-American Shu Zhang, a Columbia Business School student, and social psychologist Michael Morris, also at Columbia, which showed how a bilingual person’s brain gets hijacked by symbols, tripping up her two linguistic selves and slowing her down. Native Chinese speakers are 11% less fluent in English (in words per minute) when talking to a Chinese-American face named Michael Lee than to a Caucasian one with the same name, and produce 16% more words describing American symbols in English (White House, Marilyn Monroe, Superman) than Chinese ones (Great Wall, dragons, yin-yang). When looking at China-evocative pictures, they are also 85% more likely to use a literal translation of the Chinese word for an object rather than the English.
Cultural symbols here are like corporate logos: Picture Apple’s icon, the McDonald’s arches, Nike’s swoosh, Lacoste’s alligator, Playboy’s bunny, the Twitter bird, Starbucks’ lady, MTV’s letters or the red Solo cup you drank all that booze from in college. Doesn’t each trigger a flood of personal memories? All those papers you wrote at Starbucks; the preppy frat guys you knew (or were) in popped-collar polo shirts; the shame you still feel about those magazines you took from dad’s underwear drawer in fifth grade… The brands that invented the logos want them to sell more products. But, this study suggests, depending on the audience, they may be distracting.
The discovery that cultural cues affect language fluency so dramatically has significance beyond second-language speakers. We are all producers and consumers of speech, writing and images. We’re all trying to tell succinct stories, whether to convey news, to entertain, to educate, or sell something– and we often don’t think enough about the associations that the images we use might have. How many different places our words and images might transport our audiences, depending on where they’re coming from.
Japanese images are time-portals for me now. When I walk through the East Village past red-paper lanterns, I remember the outdoor onsen volcanic baths where my Japanese friends used to take us–a regular Japanese tourist destination, where families bathe in the river at night, by lantern light. The Ippudo ramen near NYU sends me back to the Momotaro-dori, or “Peach-Boy Street,” the main avenue in my old prefecture’s capital city, Okayama, where we used to go for noodles. Peach-Boy’s image itself is evocative: whenever I see pictures of him in sushi restaurants, walking with his dog, bird, and monkey, my mind is flooded with memories of meeting Okayama friends around the Mototaro statue in front of the eki–I mean, train station.
But to you, assuming you’re not Japanese and haven’t lived there, Momotaro’s image is just a naked Asian boy. What the new study suggests is that that blip of recognition–that jarring culture-clash of a symbol out of context–may basically clutter my communication channel with noise. These visual associative cues can communicate rich webs of content fast, but they can also interrupt fluency if they trigger the wrong chain.
“Understanding how these subtle cultural cues affect language fluency could help employers design better job interviews,” ScienceNOW reporter Emily Underwood explains. “For example, taking a Japanese job candidate out for sushi, although a well-meaning gesture, might not be the best way to help them shine.” The same lesson applies to media: displaying Japanese imagery, for example, to a Japanese viewer or reader of English will likely create unanticipated hurdles to comprehension. Picking the right chain of associations, on the other hand, in the pictures we use, might promote meaning.
Language-teachers and -learners have long known that immersion is the best and most efficient way to learn a foreign language. The Rassias Foundation at Dartmouth, where I learned my first words of Japanese in a two-week bootcamp where English was outlawed, and the Middlebury Language Schools (motto: “No English Spoken Here”) have earned their reputations on this principle. The fact that English in Japan does not tend to be taught this way, but rather with lessons on grammar and vocabulary taught in Japanese, supplemented by visits by native-English-speaking assistant teachers, may be why Japan ranks only 22nd of 54 countries in English proficiency, according to Education First.
Immersion has the obvious advantage of heavy exposure to the sounds of a language, and forced engagement. But the new study, published in the Proceedings of the National Academy of Sciences, suggests that there’s something deeper– and not strictly linguistic– about immersion abroad which facilitates fluency in a new language.
Speaking a foreign language is more social than linguistic; more like swimming or riding a bike than memorizing multiplication tables. Motivation and expressivity are key: having something you want to say, and a person you want to tell. You learn a language fast when you’re dating a foreigner, or when you’re one of two western men in a rice-paddy village, and want to make new friends. What you’re doing in those situations is marinating in associations–soaking in experience.
You talk in new ways, not just new words, when you live immersed. You make friends differently, date differently, work differently. I’d never bowed to my boss before I met the mayor of Yakage. I’d never been called “sensei” or bowed to every morning, until I met my elementary-school students. I’ll always remember my Japanese friend telling me to pay more attention to the beer in my dinner companions’ glasses, so I could refill them (no Japanese person would refill her own glass, considering it rude), and since then I’ve been sensitive to others’ bi-ru when I’m in Japan, but not in New York. The same beer glass in Osaka or New York is a different beer glass. Or as Basho put it: Even in Kyoto/ when I hear the cuckoo sing/ I long for Kyoto.
Images, whether in pictures or in words, are potent evokers of meaning. Webs of associations dependent on culture, they can animate or distract from our message–a crucial lesson for us culture-makers to keep in mind.
[Image: Flickr user Enrique Dans]