More than two decades as a journalist and communications expert has honed my radar for spotting issues that are likely to turn into industry issues – and the risk and liability associated with AI integration is a factor that is looming large. There are many cases in the news about how AI outputs may be flawed and a growing belief of an inevitable high profile AI catastrophe, but adoption of this technology continues to boom, driven by the potential of a successor to the throne of the industrial revolution.
An incident highlighting liability for me is the ‘Mobley v. Workday Inc’ lawsuit, exploring whether the AI algorithms of global HR Software behemoth Workday have discriminated against job applicants by screening out candidates based on race, age, and disability. The case is now a collective action lawsuit, despite Workday’s argument that its screening follows parameters set by the employer, thus assigning liability to the individual clients. However, as Judge Rita Lin pointed out in her ruling: “Questions remain regarding where to draw the line between someone who was and someone who was not ‘recommended.’ Does anything less than a ‘strong’ CSM [customer success manager] score qualify as a denial of a recommendation?”
At its heart, the question being asked is whether Workday essentially coded human biases into its programming, an issue that maritime organisations must consider for implementation of AI. Our sector is often focused on physical failure such as vessels running aground or ports closing, but the Workday case reminds us that liability is also baked into daily operations through the programme code. This is not a far-fetched scenario. Consider an AI bot that is used to screen orders and respond to customer queries. If that system builds customer profiles that inadvertently prioritises certain demographics or regions over others, your organisation could be accused of discriminatory trade practices.
Additionally, if we allow AI software to decide who is ‘promotable’ at sea or on shore based on data sets that lack diversity, we carry the legal liability for that bias. As AI cannot be programmed to measure soft skills, it is working with a fixed mandate set by a programmer in conjunction with the company. Thus, if the programmer is biased, the AI will also be biased – and the company will be legally accountable for that bias.
Companies must also look beyond liability to reputational risk, as there is deep-seated public anxiety regarding job replacement and the dehumanisation of the workplace. If we handle AI poorly, the result could become a conversational touchstone akin to the Titanic, Exxon Valdez or the Ever Given.
Navigating AI implementation safely requires rigorous, regular assessments of AI outputs against human judgment. We must audit our algorithms as strictly as we audit our safety ISM codes. Only by checking for innate bias through constant human oversight can we ensure that AI remains a tool for progress, rather than a collision point.
Written by Namrata Nadkarni, Intent Communications CEO
Want to get articles like these delivered straight to your inbox? Subscribe to our newsletter for insights from our roundtables and discussions with experts, alongside the key digitalisation trends and developments: https://intentcommunications.co.uk/digitalisation-newsletter/.