With its near-ubiquitous reach, artificial intelligence (AI) has infiltrated practically every industry imaginable, ranging from cybersecurity and manufacturing to supply chain management and healthcare, and, yes, even medical writing. In the world of medicine, AI perhaps is best known as a technological tool that healthcare organizations use to manage enormous amounts of information, also called big data. Often appearing in the form of software or an application integrated into a company’s preexisting software, AI allows organizations to analyze big data to extract extra information such as data trends and population characteristics that humans either lack the bandwidth to evaluate efficiently or cannot analyze organically.
Similarly, AI has gained a great deal of attention as a tool to enhance the diagnoses of various conditions such as cancer. In medical writing, organizations have come to rely on AI to assist with distilling the complex regulatory content presented into digestible summaries using plain language.
Why Natural Language Generation Matters in Regulatory Writing
Natural language generation (NLG) is a specific type of AI that converts technical jargon into plain language, but the technology is not exclusive to regulatory writing—or medical writing, for that matter. In fact, many industries employ NLG to help them connect with their target audiences, as seen in journalism, business writing, and financial writing, in addition to medical writing.
While the concept of NLG does not frequently surface in areas of AI that do not involve content creation, it entails some of the same fundamental principles found in other areas in which AI is used, such as machine learning. Recognizing how machine learning works is imperative to understanding how AI applications enhance various work-related tasks and the value that lies in their implementation.
Machine Learning Is Critical to AI Success
Also known as predictive analytics, machine learning is a critical component to AI integrated into any pre-existing software or program. By basic definition, machine learning is the phase during which AI essentially “learns” how to do its job by being trained to recognize certain patterns from which it develops a problem-solving algorithm.
During the learning phase, the software is exposed to hundreds of pieces of data relevant to its respective field or task so that it essentially “learns” what to recognize. Using diagnostics as an example, AI used in dermatology can be “taught” to diagnose melanoma by being exposed to hundreds of images of melanoma in various stages, shapes, sizes, and skin tones to facilitate the application’s ability to identify cancerous lesions with high fidelity.
NLG employs a similar process to develop its algorithm in regulatory writing.
Navigating the Intricate Pitfalls of AI
As with any man-made invention, AI in medical writing is subject to limitations, challenges, and the potential for error. For starters, the information used to train AI during the learning phase ultimately determines the technology’s level of fidelity and reliability. For example, while one kind of melanoma, called superficial spreading melanoma, accounts for 70% of all melanoma, exposing AI only to superficial spreading melanoma and not other types of melanoma could result in missed diagnoses.
The subject of melanoma diagnoses also brings to attention the differences in which the disease presents in diverse populations, as melanoma presents differently in darker-skinned individuals than it does in people with lighter skin. While people with darker complexions, such as Latinos and African Americans, have lower incidences of melanoma than Caucasians, melanoma is frequently diagnosed in more advanced stages in these particular populations. For this reason, engaging technology programmed to recognize the condition across diverse populations can help improve overall outcomes.
On a smaller scale, some organizations still struggle with using the technology in a way that helps them optimize big data. In such cases, some organizations seek the help of software management firms to better manage their data and customize the information they collect from that data using AI to suit their individual business needs. In other words, such firms help corporations assign meaning to tremendous amounts of data that may otherwise be useless. However, despite significant advancements in AI, some experts still say that the full potential of big data remains untapped.
AI Meets Resistance, Doubt in the Workplace
Understandably, integrating AI into the workplace has caused some angst in professionals across industries where it is used. Some automatically discount the technology, questioning the accuracy and reliability of the information it provides; others fear it may threaten job stability. For example, many physicians expressed their concerns that the technology might replace them after some studies showed that AI led to more melanoma diagnoses than physician-initiated efforts.
Similarly, some professionals in regulatory writing question whether NLG can do the job as well as a human—especially when the writing requires technical skill coupled with varying degrees of artistic nuances. While there are still certain areas of writing that benefit from human skill, AI can improve efficiency with regard to some of the more time-consuming clerical tasks such as document or content removal and redacting text.
Forecasting the Future of AI in the Workplace
Clearly, engaging technology at higher levels brings certain questions and concerns to the forefront. While no one holds a crystal ball, perhaps one can safely assume that AI will not completely replace human resources any time soon. After all, there still are—and always will be—certain roles and fields that require a human touch. Writing is one of those fields. Even the best technological simulations cannot recreate the energetic, empathetic experience that occurs as a positive consequence of human involvement and interactions. However, engaging AI, and more specifically, NLG to handle some of the more clerical aspects can only enhance efficiency and productivity.
Ultimately, when used with care and understanding of its utility and limitations, AI can enhance human performance, quality, and productivity. In other words, AI does not offer a replacement for human talent but rather an additional tool or adjunctive resource in one’s arsenal for success.
An education session on artificial intelligence and natural language generation will be presented at the 2019 Medical Writing & Communication Conference in San Diego.