Allyson Kapin on Future AI in Social Media

Maci Peterson of On Second Thought, won the Women Startup Challenge Pitch Competition 2015, and received a check for $50,000 from Women Who Tech, presented by Allyson Kapin9 (right). Image: Kristin Johnson, Women Who Tech

Interviewing really bright people about their perspective of the business has to be my favorite part of writing a book. For Welcome to the Machine, I decided to reach out to Allyson Kapin, Co-Founder of Rad Campaign and Founder of Women Who Tech, to discuss how AI will impact social media marketing. Here are her answers to my questions.

GL: What value does AI offer the social media world?

AK: One of the biggest opportunities AI offers the social media world is that it can tailor content based on what people search, like, and purchase in real-time. Marketers can use the data to create better audience/customer personas.

The more AI learns your habits and preferences, the better it gets at serving up content and news that will interest you or products that you’d be interested in purchasing because it has a deeper understanding of your views, tastes, styles, and preferences. This is especially important for marketing consumer businesses, nonprofit advocacy, and political organizations and campaigns.

Social networks have heavily invested in AI. For example, Twitter categorizes every single tweet of yours so that it can serve up tweets based on your interests so you stay on the platform longer reading and retweeting more content. More time spent on their platform means the more you get hooked on Twitter and the more Twitter can sell ads and target you.

While targeted content based on your interests can be helpful because it helps people quickly filter content that interests them, AI is also contributing to a vicious cycle where users of social networks are living in a filter bubble and that can be dangerous.

GL: Can it further impact content and curation beyond the large social network streams?

AK: Yes, AI is becoming more affordable for businesses to integrate. It’s not just for the large social networks or tech giants. For example, more and more AI is being used by businesses for customer service inquiries.

AI-powered chatbots are helping businesses of all sizes respond more quickly to customer service requests and improve the customer experience and create more operation efficiencies. While it’s still an early technology that has kinks to work out, it’s the future of customer service and as it grows, it will become even more affordable even for the smallest local businesses.

In addition, we’re seeing more open source integrations with services such as with Google AI and that will only continue to grow and be more adapted.

GL: How can AI or machine learning be useful in community management?

AK: For community management, social networks like Facebook and Twitter have been using AI to either help flag inappropriate content or when inappropriate content is submitted for review, AI scans the content and determines if it’s inappropriate.

The problem is that AI is not advanced enough to understand human context in conversations. It does not understand different cultures or nuances. AI and ML can easily be biased as it’s often coded from a very white, male, and US/European centric perspective as that’s who’s been funded to develop most of the major AI technology.

Here’s a good example to illustrate my point. About 45% of the images from ImageNet, (a widely used image database) is from the US. Other countries who have much larger populations than the US (India and China) are barely represented. So when computer vision algorithms looked at a photograph of a traditional US bride dressed in white it labeled it as “bride”, “dress”, “woman”, “wedding”, but when it looked at a photograph of a North Indian bride it labeled it as “performance art” and “costume.” #Fail

GL: Can AI protect the user experience?

AK: Yes, AI can definitely help protect the user experience and Spirit AI, who was one of our Women Startup Challenge winners is one of the best examples I know of. The founding team witnessed significant harassment happening in the online gaming community, especially targeted at women players.

Their response was to create AI to help gaming communities better integrate community management and help detect, evaluate, and respond to various forms of harassment and issues players were experiencing in real-time. As the AI continues to be implemented across more games and gathers more data on players and how they interact with each other, AI can also help predict issues before they escalate.

GL: What are the ethical challenges of using machine learning AI?

AK: There are several ethical challenges in AI and ML that keeps me up at night. These are challenges that are not easy to overcome and the tech world needs to do better at addressing these ASAP before it’s too late.

1. Over Collection of Data: AI thrives off of data. The more data it receives and processes, the more it learns from the datasets. But companies are collecting a lot more data on people then they need to and that’s worrisome.

2. Data Security: Data is not secure. We need much stronger data security standards and standards that are enforced. There have been way too many major data breaches, with little penalties, and it’s unacceptable.

3. Filter Bubbles: People can get trapped in a filter bubble where they only see content, news, and perspectives that they believe. This is incredibly dangerous for media literacy and democracy.

4. Consumer Manipulation: This dovetails off of living in a filter bubble. AI can and in many cases is built to manipulate people into doing things. I.E. buying a product they don’t need. This contributes to wasteful spending and accumulating more “stuff” this planet does not need and will probably end up in landfill. On a political scale, it can manipulate democracy and elections as we have witnessed.

5. Biases: AI is prone to a lot of biases across all industries and it can cost people their life. For example, last year researchers used ML data to identify skin cancers from photos. The dataset it used contained photos of 95% white people and only 5% people of color and the algorithm was not tested on people of color. That’s a huge bias that could cost people their life.

GL: What role do humans have to play in the AI social media future?

AK: Humans play a huge role in the future of AI. People creating the AI have the power to get this right and think through and overcome the ethical challenges. Companies are in such a rush to develop “the best” AI and make a lot of money, but they are not taking the time to think through all of the ethical challenges it brings to the world. And that needs to change! In addition, governments around the world need to become better educated on AI and how it will impact not just their country but the entire world.