OceanGate’s Titan submersible tragedy risk assessment by AI

Risk management is the systematic process of identifying, assessing, prioritizing and mitigating risks to minimize their potential negative impact on an organization or project. It involves analyzing potential hazards, vulnerabilities and uncertainties that could arise. This is done to develope strategies to address and reduce the likelihood of consequences and their size.

In a broader sense, risk management aims to increase the likelihood of achieving objectives while minimizing the potential for harm or losses. It involves a structured approach that includes risk identification, risk analysis and evaluation, risk treatment and ongoing monitoring and review.

Risk management is an essential aspect of decision making that extends beyond organizational or project settings. Its principles find relevance in the everyday lives of individuals. In personal life, people encounter uncertainties and risks regularly, prompting them to make conscious decisions to mitigate potential harms. By identifying, evaluating and taking proactive measures against risks, individuals make informed choices that prioritize personal safety and security.

What’s the difference between a human-guided risk assessment and an AI generated risk assessment? A human-guided risk assessment involve human expertise, judgment and subjective considerations, allowing for flexibility, adaptability and effective interpretation and communication, yet biased and limited to ones knowledge and level of understanding. AI-generated risk assessments rely on data-driven analysis, consistency, scalability and reduced bias. Not every #AI generated risk assessment is equal, as not all AI function in the same way.

In this experiment, to better understand how an AI would operate such request, OpenAI #ChatGPT is prompted to assess the risk of submersible expedition by #OceanGate. The premises of the experiment is that the AI has only access to the information prompted and does not know the sequence of events and how the tragedy unfolded.

This is the prompt given:

This is the answer the system generated:

It’s fascinating how, Artificial Intelligence picked as three options to go, scientific and human-knowledge related topics. Instead, there is no indication to an emotional value added as life-changing experience or related feelings of exploration and satisfaction.

Then we move to the reasons not to go:

Interestingly, in the risk assessment, the safety concern places first. The cost is not mentioned, being instead a considerably constriction for a person.

Proceeding with the experiment, more information is given, to have a more informed risk assessment:

Based on new information the first answer that the AI generates is still a safety concern:

Followed by a series of different topics that are often cited in risk assessments’ frameworks:

A final prompt is given:

The AI generated the following overall evaluation, leading as for the first set information an overall answer to go, and based on the second information not to go on the trip.

Interestingly, this experiment led to an understanding on how structured is a risk assessment led by a non-human entity. Especially significant is the fact that the system is not biased with emotions, and cannot be classified into "naturally inclined to" or "naturally adverse" to risk, as humans. More research is necessary as this quick experiment is not statistically significant, nor able to represent a structured framework of how AI generated risk assessment works.