It's no longer a question of *if* inclusive content performs better (it's been proven to do just that across industries, time and time again). The question now is - is *my* content inclusive? Now, you can quantify and measure inclusion responsibly. Alltold was founded by Responsible AI leaders who worked on this together in Google's AI Research group. At Alltold, they've partnered with experts and advocacy groups to ensure a responsible approach from day 1. And they have a singular focus on making inclusion data easy to use. When content is inclusive, both businesses and society benefits.
Alltold的外部链接
Alltold转发了
?? Paul Hiebert at ADWEEK for covering important topics and using data to tell the story. Alltold evaluated skin tone representation in the top 100 viewed ads during Black History Month over the past 5 years for this article. Link to read in the comments. Thanks to our partners Innovid for sourcing the ads (thanks Dan Mouradian!), and Melissa Rosenthal and Jess M. at System1 for bringing the performance data to the table.
Alltold转发了
The Super Bowl is such a unique TV event. At a time when our media options are vast, diverse, and on demand (in stark contrast to the TV options when I was a kid), the Super Bowl is a time where most of the country sits down and watches the same thing at the same time. I'm always curious to see how advertisers choose to reflect (or not reflect) the diversity of the viewing audience in their Super Bowl ads. If you've been here for a minute, you know we've analyzed representation across 50+ years of Super Bowl ads, and this year is no exception. Although Super Bowl Sunday is always a late night for us, I always love running this analysis. Cecilia Garzella wrote a thoughtful and detailed piece using Alltold data showing who we saw in the Super Bowl ads this year for Campaign US. And a special thanks to our friends System1 for bringing the customer reaction data to the table. Article link in comments #AI #ResponsibleAI #SuperBowlAds
CEO Morgan Gregory shares some of the findings of our latest benchmark report in a Beet.TV interview. Watch for more to come with our partnership with Innovid!
????? Inclusive ads aren’t just good ethics—they’re good business! Discover how inclusive content drives real results in this Beet.TV interview with Alltold CEO Morgan Gregory ? https://ow.ly/1VEi50U79M9 #AdTech #Inclusivity #AI #Innovid
The most surprising finding in the report we just launched? That we’re at parity for feminine : masculine gender expressions (what?!?!) in the top viewed ads of H1 ????? My take: the efforts, the advocacy, the movements, the initiatives - they’re working. When we put a focus on an aspect of inclusion, we can move the needle in a meaningful way. ?? AND there’s still room to grow. ? We see relatively few people on screen who show as gender non-conforming. There’s an opportunity to bring these expressions into the spotlight. When we intersect gender expression with another identity attribute, we can see more nuance in the data. ? We see more people with a feminine gender expression under age 30 than above. ? We see more people with a masculine gender expression with the darkest skin tones on the Monk Skin Tone scale. ? We see that people with smaller body sizes are more often those with a feminine gender expression. ? And when we look at the numbers by vertical, we see that there’s quite a large range. If you were to guess, which bars in the graph below do you think belong to: ?? CPG: Food & Bev ?? CPG: Household ?? Home & Garden ?? Pharma ?? QSR No cheating! JK - download the report to see. Answers on page 5. Link in comments. Again, big thank you to Innovid, Dan Mouradian, Megan Garnett Coyle, and Tameka L. for your partnership in this work!
Since day 1, we've had a goal to make benchmarks for inclusive content open and available to everyone. We started with Super Bowl ads, and today, we're excited to share benchmarks that look at ads that are run - not just around tentpole events - but 24/7/365. We're thrilled to partner with Innovid to bring you this view of the state of representation in the most viewed CTV ads of H1 2024. We chose this set of ads because they represent tens of billions of ad views - and have the largest exposure surface area. As viewers, because of this exposure, they're the ads that most strongly influence our beliefs about the people and world around us. Check out the report, and let us know your thoughts! https://lnkd.in/dh4jd8d6 Don't miss the Q&A at the end of the report with CEO Morgan Gregory and Dentsu's Cultural Fluency expert Tameka L.. And as always - get in touch to measure your ads for inclusion with Alltold's *responsible AI* solutions #ResponsibleAI #RepresentationMatters #InclusiveMarketing Innovid Dan Mouradian Tameka L. Megan Garnett Coyle
As technology pioneers in inclusion and representation analytics, we've had a responsible approach from day 1, throughout our entire technology platform and business.
?? Not all analytics are created equal. ?? ESPECIALLY when it comes to analytics that deal with subjective measures of identity. Take Gender Expression. Before we built Alltold, all of the reports and analytics you would see on the market would use Gender, and would use binary options: Male / Female or Man / Woman. As Dr. Joy Buolamwini pointed out in her book Unmasking AI (which I highly recommend), “Having the ability to define classification systems is in itself a power”. The responsibility of that power is something we do not take lightly. When we built Alltold, our first order of business was to build relationships with experts and advocacy groups in order to collaborate and build responsible and respectful classification systems. We had 2 firm starting points when it came to Gender Expression: 1. We wouldn’t measure Gender or Gender Identity. Gender is an internally held identity attribute, and you can’t tell someone’s gender by looking at them. Often, in the content we analyze, visual cues are all we have to work with. 2. We didn’t want it to be binary. We recognized that we couldn’t build a Gender Expression scale that mapped to the diversity of genders that exist in the world (yet?), but we did want to create space for something that fell outside of what we typically see for feminine and masculine. After multiple conversations and iterations with Meghan Bartley and Tristan Marra at GLAAD (thank you again!) we landed on Feminine, Masculine, and Gender Nonconforming. That’s what you see when you view our reports. But what you don’t see is the detail in how we describe these, the reference images we jointly agreed on, the training we built for our annotation team, and the rigorous processes we have in place to ensure accurate, high quality output. Let’s say another company were to take our scale. They could build reports that might look like ours. They’d use the same words. But that doesn’t mean that they have implemented the scale responsibly, or that the end results are accurate. And this is just one of the identity attributes we look at. There’s so much nuance to inclusion analytics, it’s critical to have a thoughtful and responsible approach. Let me know where you've seen #responsibleAI in action ??
??Inclusion=Income?? Our friends and partners at the Unstereotype Alliance released a new report yesterday that (again) proves that Inclusive Advertising drives bottom line financial results. Their findings? Brands with more inclusive advertising: ? have 3.46% higher short-term sales ? have 16.26% higher long-term sales ? are considered trialed more ? have higher customer loyalty ? are valued greater by consumers ? have higher brand equity They found these to hold across geographies and product categories. We couldn't agree more with their recommendation to make inclusion commitments measurable. At Alltold, we've built responsible technology that can quantify and measure inclusion in ads quickly, cost effectively, and at scale. Check out the report here: https://lnkd.in/ei6Q7q_W
Alltold转发了
How do image generation platforms generate people? I ran a very non-scientific experiment to find out. LLMs and Generative AI platforms are trained on an enormous amount of data - think: the entire, open, scrape-able internet. These large models learn the patterns (and biases) present in their training dataset. At Alltold we build responsible AI to point out those biases and stereotypes - both in content created by people, and by AI. I have a list of other test to run, but need help picking the platforms to test. What image or video generation tools do you use?