Predicting the future has a knack for making even the smartest among us look foolish. Remember when tech pundits in the ’90s scoffed at the internet, dismissing it as a passing fad? Fast-forward to today, and our lives are intertwined with the web in ways they could never have imagined. In this post, we’ll explore another frontier that’s set to revolutionize our industry: the role of artificial intelligence (AI) in both creating and analyzing SEND data. Buckle up as we dive into the potential game-changer that AI promises to be for nonclinical data exchange.
Did that sound like me? That was what ChatGPT produced after I asked it to read my blog and then write me an opening paragraph for a post to discuss the possible uses of AI in both creating and analysing SEND data, using my typical writing style. I think it did a pretty good job, although I don’t think I’ve ever used the word ‘intertwined’ before. Maybe I should start?
AI seems to be all around us these days and would appear to excel in data analysis and pattern recognition, natural language processing, and image and video recognition. It can automate repetitive tasks, and it’s used in applications such as recommendation systems, and predictive analytics. AI has clear uses in fraud detection, cybersecurity, and supply chain management, and seems to be used in innovative applications across various industries. I’m sure many of us will have used the chat facility on a website and happily found ourselves getting assistance from an AI bot.
So, does it have an application for SEND? I have heard someone flippantly state that AI should be able to convert study data to SEND instead of people. Personally, I think that’s nonsense (at least in the short term), but I think we should be able to use AI as if it were our own highly efficient assistant, taking on repetitive tasks and making intelligent suggestions for things like how to resolve conformance issues in our data.
I say this because I don’t believe that we should trust AI to make decisions, but rather use it as a tool to assist us. For example, when generating a SEND dataset, let’s say that there’s a certain term used in data collection that needs to be translated to SEND Controlled Terminology (CT). We would not want AI to make that decision for us, but wouldn’t it be great if the application could detect this issue and then suggest the most appropriate CT for us to choose from?
Another possible use could be in extracting trial design parameters and other metadata from the textual portion of the report as well asa protocol document, SOPs and other supporting documentation. AI’s Natural Language Processing is capable of understanding, interpreting, and generating human language, (which is used in applications like chatbots) and so should be able to read documents and extract the relevant information in a fraction of the time it takes a human. The human would still need to check and accept such suggestions, but it should get us to an answer much quicker.
For anyone who’s responsible for creating SEND datasets, I don’t see AI taking your job away any time soon, but it could make it easier. And with that thought I ran out of words before we could start discussing the role AI could play in searching, summarizing and analyzing our standardized nonclinical data. Maybe that should be the topic of a future blog.
‘til next time
Marc
