The Role of Machine Learning in Public Relations

Don't miss all the great posts on our partner network, NATIONAL PR's Bold Thinking blog, where this post originally appeared.Machine learning is already transforming industries around the world, offering significant cost savings and efficiency gains, while opening up new opportunities for forward-thinking companies to leapfrog competitors. How will machine learning affect the public relations industry?

What is machine learning?

Let’s first begin by defining what machine learning is. Machine learning is a capability of machines to learn how to perform or improve the performance of a task without explicitly being programmed to do so.In the early days of computing, machines needed us to tell them exactly what to do. They had no independence, no ability to deal with any exceptions to rules. As the field of artificial intelligence progressed, computer scientists developed ways, mostly through statistics and mathematics, for machines to learn how to deal with irregularities and errors without having explicitly defined rules for every possible error. This is the foundation of machine learning as it is today.In the early days of computing, machines needed us to tell them exactly what to do. They had no independence, no ability to deal with any exceptions to rules. As the field of artificial intelligence progressed, computer scientists developed ways, mostly through statistics and mathematics, for machines to learn how to deal with irregularities and errors without having explicitly defined rules for every possible error. This is the foundation of machine learning as it is today.

Kinds of machine learning

Machine learning technologies fit into two broad categories today: supervised and unsupervised learning.Supervised learning is when we give a machine a needle to find in a haystack. We tell it what the needle looks like, then set it to the task of digging through the haystack. We use supervised learning for tasks like recognizing when a brand’s logo appears on Instagram, or what kind of sentiment surrounds the mention of a client’s name in a news piece.Unsupervised learning is when we give a machine an enormous pile of data and ask it to sift through it, then categorize its findings. Suppose, for example, we had all the social media conversations about a client aggregated into a large database. How would we ever sort through the millions of social media posts? We couldn’t, we’d run out of lifetime. However, an unsupervised machine learning process could sort through it all and identify clusters of content together. One cluster might be customer service issues. Another cluster might be competitors. With unsupervised learning, we’d be able to reduce the sheer volume of content down to something that we humans can understand and take action on.

How machine learning impacts public relations

Consider how these two categories of machine learning will impact public relations. How much of our work is based on recognizing and classifying data?For example, when we do coverage tracking, we’re manually performing supervised learning tasks. We may use software to identify coverage of a client’s name or products in bulk, but then we manually sift through as much coverage as we have staff to support and produce reporting from it. This is labor-intensive and not a lot of fun for the staff who have to do it day in and day out. Machine learning could automate a substantial portion of this drudgery, freeing up staff to be more creative or spend their time obtaining new coverage, rather than reporting on past hits.Additionally, with the explosion of new content and news daily, it’s more and more likely that on any given day, we miss coverage of our clients due to sheer volume. This year, we average 200,000 news stories per day. We send 42 billion messages per day on WhatsApp. We send 18 billion text messages and read 59 billion email messages a day. We post 415 million tweets, 80 million photos on Instagram, watch 3 billion videos on Facebook, swipe 1 billion times on Tinder – all in a single day. With that media backdrop, there’s no question we’re missing coverage and conversations about our clients.Machine learning – particularly supervised learning tools that can recognize our clients – will help ensure we miss fewer important individual communications while helping us understand the communications landscape in aggregate.The top question on everyone’s mind about machine learning and artificial intelligence is, will these technologies replace us all? The answer in the short-term is no. Individual tasks – like coverage scanning or reporting – may be streamlined and reduced to the push of a few buttons. However, a key part of our message as “Trusted Partner” is the human to human relationships we have with our clients, our media sources, our influencers, and each other. No machine will automate those relationships in the foreseeable future. As long as we continue the second part of our mission – Bold Thinking – and find ways to think boldly, augmented by our machines, we will continue to provide great value to our clients, build their businesses and ours, and grow our relationships across the industry.