Decoding the Social Effects Of Media with Machine Learning

What if media had been optimized to profit individuals? This thought-provoking query is on the core of Concord Labs‘ mission. A nonprofit group headquartered in New York Metropolis, Concord Labs strives to raised perceive the affect of media on society, and construct communities and instruments to reform and remodel media techniques.

As Brian Wanieswki, Govt Director at Concord Labs places it: “The media techniques that we’ve now, for higher or worse, have turn into outrage machines and sorting machines that put individuals into teams of like minds. The enterprise incentive constructions of those techniques are such that the extra outrage there may be, the extra revenue there may be. Political occasions internationally lately have borne out what these media techniques produce, and it’s actually fairly poisonous, and fairly laborious to get something carried out inside. There are every kind of pure divisions between individuals, however these media techniques have a tendency to strengthen these divisions. So, the primary query that we’re asking is, What’s the scope of this downside? After which, What can we do to resolve it?”

Concord Labs use knowledge science and machine studying to reply these questions. Ranging from consumer surveys and media knowledge, they developed superior pure language processing pipelines that may establish how social points are represented in media, how they’re consumed by totally different audiences, and how much affect that consumption has.

So how do you get that knowledge in first place? Brian says: “All of the media knowledge that we want lives inside non-public corporations. We knew that knowledge sharing can be central to our mission, and this is the reason we’re structured as a nonprofit. We’re working within the public’s curiosity in a non-partisan manner. We’ve carried out large knowledge sharing offers with giant corporations, in addition to startups scraping totally different corners of the media ecosystem that we’re eager about: web TV, web radio, and so forth. We’ve got about 10 knowledge companions in the mean time, and we’re at all times seeking to broaden.”

Because of their knowledge companions, Concord Labs has collected over 50 Terabytes of numerous media: TV, net, cell, tune lyrics, closed captions, social media, and extra. This undoubtedly matches the definition of huge knowledge (quantity, velocity and selection). Working with languages like Golang, Python and R, the Concord Labs knowledge science and engineering groups depend on AWS providers akin to Amazon Aurora, Amazon Athena, AWS Glue and Amazon Elastic Kubernetes Service (EKS) to construct their knowledge ingestion and processing workflows.

As soon as the information is in-house, Concord Labs make it protected, safe, and accessible to a community of educational researchers who use it to analyze the affect of media techniques on politics, society, and tradition. Laura Edelson is one in all these researchers. A Ph.D. Candidate in Pc Science at NYU’s Tandon Faculty of Engineering, she research on-line political communication and develops strategies to establish inauthentic content material and exercise. Concord Labs supported her on the Advert Observatory undertaking, an exploration of political adverts on Fb.

Concord Labs additionally work on their very own initiatives, such because the Narrative Observatory. “A story is a narrative sample that recurs throughout totally different sorts of tales and media. You’ll discover them in tune lyrics, TV reveals, information articles, and extra“, says Brian. The Narrative Observatory helps establish narratives on explicit matters and observe them over lengthy durations of time, and throughout totally different media sorts.

With preliminary funding from the Invoice & Melinda Gates Basis, Concord Labs studied narratives linked to the subject of poverty and financial mobility in america. Gathering thousands and thousands of paperwork (on-line information, social media, music), they first recognized the principle narratives current in media. Then, utilizing segmentation strategies, behavioral knowledge on over 50,000 Individuals and surveys, Concord Labs outlined 4 audiences, in addition to their dominant narrative, their core values, and their views on particular social points. Lastly, Concord Labs studied how every viewers consumed narratives.

To allow funders, companions, and media corporations to achieve a deeper understanding of the cultural areas their audiences occupy, they constructed a captivating web site, obiaudiences.org, the place you possibly can decide an viewers and examine the related media feed. In different phrases, you possibly can actually see the world in another person’s eyes: what points they care most about, what media they learn most, and so forth. This helps to know the perceptions that totally different individuals have on sure points, and as Brian places it: “When you’re making an attempt to achieve individuals, it’s essential to know the media world they inhabit, and what may be really related to that world.”

Not too long ago, Concord Labs led a undertaking funded by the Mozilla Basis on defining a wholesome narrative for synthetic intelligence (AI). Finding out the TV consumption habits of over 80,000 US adults, and connecting them with closed captioned transcripts and adverts, they recognized and named the principle media narratives on AI. Every narrative features a definition of AI, the emotion that it creates in individuals, and whether or not they suppose that AI will result in a cheerful or an sad ending.

Concord Labs recognized 4 major narratives on AI. Two are extraordinarily detrimental and fear-inducing. “Device of Tyranny” says that AI shall be utilized by governments to oppress individuals. “Robotic Overlords” says that we’ll by no means be capable of management AI, and it’ll find yourself ruling us. On the different finish of the spectrum, the “Needs Granted” narrative is extraordinarily constructive: certain, we don’t perceive AI, however it’s a magic wand that may clear up all our points. The final narrative, “Augmented Intelligence”, is extra balanced: sure, AI is a good alternative to enhance our every day lives, however it’s additionally able to being unfair and even harmful. We’re answerable for designing it, controlling it, and ensuring it’s used to assist us, to not harm us.

Concord Labs discovered that the “Needs Granted” narrative was probably the most outstanding (67%). It shines a constructive mild on AI, however its naive and over-optimistic imaginative and prescient can cover the legit questions that AI raises. Nonetheless, it’s an excellent place to begin to have interaction audiences, educate them with the “Augmented Intelligence” narrative, and improve their consciousness on each alternatives and challenges.

Closing this submit, I’m questioning which AI narrative I’ve really promoted right here, willingly or not! What do you suppose? One factor is definite: Concord Labs is utilizing AI to assist us perceive how media influences us each day, and the way we are able to create a extra democratic society. That is essential work, and we’re humbled that they picked AWS to assist them attain their objectives.

For extra data on Concord Labs, please go to harmonylabs.org and harmonylabs.medium.com.

– Julien