It’s reasonable to feel cautious or even trepidatious about AI. Technological revolutions are often disruptive, and we don’t have to look too far into the past for a recent example: the rise of the internet and the advent of e-commerce changed how we interact with one another forever, and reshaped the towns and cities we live in. But behind every piece of content and at the end of every online interaction there is still a human being – and the same will ultimately be true of responsible AI.
In this article, we’ll explore how the ‘human side’ of AI helps the technology remain safe, ethical, responsible and effective in life science and beyond.
The human side of AI development
“AI systems are the product of many decisions made by those who develop and deploy them.”
AI tools are not yet capable of independent thought. In fact, they may never be. Instead, they’re the product of human data and human programming – one of the reasons they might feel frightening or uncanny is that they often reflect our own biases and prejudices back at us. The flip side of this is that the development stage gives us an opportunity to dictate how AI systems work: how they process data, how they generate responses, and the rules and standards they abide by.
This is the idea behind ‘responsible AI’ – a commitment to building AI tools that adhere to ethical and societal values. Under responsible AI, systems are developed with human rights and ethical standards at the top of mind. Guardrails – rules and filters designed to mitigate risks and minimize ‘hallucinations’ – ensure these standards are enforced. So in a very real sense, AI development is a human process designed to best serve humans.
If you’d like to learn more about our commitment to responsible AI, read our eBook here.
AI systems are the product of many decisions made by those who develop and deploy them.
Here at Within3, we use another layer of human input when training our Launch Intelligence™ AI models. We recognize that factors including nationality, faith, age, gender, sexuality and even upbringing can dictate what ‘fair’, ‘true’ and ‘accurate’ mean from person to person. That’s why we use a diverse panel of experts from a wide range of backgrounds to rank and score our training data, so our AI outputs aren’t representative of any single point of view. In this way, we can eliminate unconscious bias as much as possible.
“We have a committee of people that come from different educational, experience, ethnic, and industry backgrounds,” explains Within3’s Chief Technology Officer, AI and Analytics, Jason Smith. “They don’t have to agree on things – we want that friction.”
Furthermore, we continue to rely heavily on human input and expertise as our AI models are developed and used. Our models are thoroughly tested for bias and fairness before deployment, and audited regularly, while our outputs are continually monitored for accuracy and relevance.
The human side of AI insights reporting
AI is helping to automate the process of insights reporting – slashing the time spent on this critical activity from months to minutes. However, the process will never be fully automated – human analysis and oversight will always remain a crucial part of the equation.
You can simplify the process of insights reporting to look a little like this:
Data
↓
Insights
↓
Action
First, data flows into the organization from numerous sources. There’s social data, claims data, field medical data, congress reports and much more. This data is structured and unstructured, fragmented, and likely encompasses a variety of formats from spreadsheet entries to free text and everything in between. Looking for actionable insights within this data is worse than looking for a needle in a haystack – sometimes you’re looking for a needle, sometimes it’s a thumbtack, and of course there’s more than one stack to search through. This is the process that used to take months. Launch AI can do this very quickly and efficiently – cleaning and integrating both internal and external data before surfacing insights that are directly related to your specific launch strategy. But it can’t do the next part alone.
Within3’s Launch Intelligence™ is capable of providing recommended courses of action based on the available data, but it can’t take action. Nor should it. This is where human input and human expertise is so invaluable. AI arms you with knowledge; you decide what to do with that knowledge, and how to execute it. Jason Smith describes launch AI as a “combination of processes, humans, and technology that come together both in sequence and in concert.” Putting the human back in the loop at this stage is crucial to ensure trust is maintained and that our outcomes are as accurate and relevant as possible. Human oversight allows our models to be tweaked and adjusted over time – enabling us to refine outputs and prevent drift.
“Reporting is rarely anyone’s full-time job,” points out Samantha McAloney, Senior Vice President of Product Management at Within3. Launch AI is not about replacing human beings, but empowering them to make insights reporting more efficient, accurate, and successful. “The more AI takes on that burden, the more people can get back to doing the work that really matters,” Samantha explains.
Human beings are at the heart of everything we do here at Within3 – from developing our AI models to assessing their outputs, and refining the process for future iterations. After all, when you’re talking about the launch of a new therapy, “you’re talking about someone’s life, someone’s treatment plan,” says Jason Smith. So “it’s all well and good trying to do cool things with the data and the technology, but at the end of the day we have to remember why we’re here.”
If you’re ready to discover the human side of AI, book a demo of Within3 Launch Intelligence™ today.
