ChatGPT essay series | A Statistician’s Perspective: a reliable ally and a potential opponent that looks like Wikipedia

After collecting contributions from experts in a variety of fields, the last one in the series comes from a statistician. Antonello Maruotti (Professor of Statistics at the LUMSA University of Rome) discusses the potential of ChatGPT to revolutionize teaching, academic research, and consultancy.

Abstract: On the academic side, the text highlights that tools like ChatGPT can enhance education, while human expertise and judgment remain indispensable in complementing and contextualizing the generated outcomes (which are sometimes claimed to include graphs, charts, and visualisations when ChatGPT is actually only capable of generating text). The system can automate routine and repetitive tasks (a good example in the text is that of calculators applied to the learning process), but cannot replace the role of a teacher or lecturer, who serves as a mentor, role model, and provides guidance.

On the consultancy side, the impact might be more relevant: ChatGPT can be a reliable ally, but comes with limitations and risks (e.g., jobs losses). While ChatGPT can consolidate existing textual knowledge and make it easily accessible, it falls short in critical aspects of data analysis such as critical thinking, strategic planning, and complex problem-solving. It is crucial for consultants to adapt with new skills and seize the opportunity presented by time and efficiency gains, to achieve a balance between human factor and automation.

Full text:

ChatGPT, and Large Language Models in general, are highly impressive and have the potential to revolutionize the way we approach teaching, academic research, and consultancy. However, the rapid advances in artificial intelligence may raise concerns, especially for those involved in teaching data analysis or introducing advanced methods in this field. The fear of data analysts becoming obsolete is not unfounded, especially with the viral emergence of ChatGPT, which has captured the public's imagination and showcased the remarkable capabilities of artificial intelligence. It is important to note that ChatGPT is not infallible. Despite its widespread popularity, the website itself acknowledges that the model sometimes produces plausible-sounding but incorrect or nonsensical answers. There have been instances where it has made glaring mistakes or referenced non-existent scientific studies. Therefore, while ChatGPT can be an invaluable tool, it also comes with limitations and risks. In conclusion, ChatGPT can be both a reliable ally and a potential adversary. It is crucial to approach its outputs with caution, ensuring critical evaluation and verification of the information it provides. The human expertise and judgment remain indispensable in complementing and contextualizing the capabilities of AI models like ChatGPT.

As your trusted and real human friend, it's important to provide a realistic perspective on ChatGPT's capabilities. While it may come across as assertive and talkative, promising to solve all your problems like Mr. Wolfe in Pulp Fiction, the reality is somewhat different. While ChatGPT can assist you in data analysis, it tends to offer more of a dream-like portrayal rather than providing solid and reliable analyses. It's true that ChatGPT may seem capable of accomplishing the impossible, offering solutions to all your data analysis problems like a modern-day genie in a bottle. However, the truth is that its capabilities are limited. For instance, ChatGPT falsely claims that it can generate graphs, charts, and visualizations when it is actually only capable of generating text. While we can't predict what future versions may bring, it's important to recognize the current limitations. In its current state, ChatGPT can support you with several basic tasks. It can assist in performing exploratory data analysis by providing summaries and advice on patterns and relationships in your data. It offers suggestions on statistical techniques for further analysis (although they may not always be appropriate). It guides you through the data analysis process with predefined steps, lists useful algorithms for data preprocessing and analysis, and provides assistance in writing code for various software programs (although it may not always work perfectly in every situation). While these features are valuable and convenient, it's essential to remember that ChatGPT is a tool that consolidates existing knowledge and makes it easily accessible to general users, much like the role Wikipedia played two decades ago. However, it does not replace the need for a solid understanding of underlying concepts, result validation for accuracy and reliability, and human intervention. Furthermore, ChatGPT falls short in critical aspects of data analysis such as critical thinking, strategic planning, and complex problem-solving. It does not possess the ability to help users fully comprehend concepts or effectively implement them. Its suggestions and analyses are basic and still require human intervention. Therefore, individuals working in academic research, introducing novel methods, and employing a mix of analytical, computational, and data analysis capabilities are unlikely to be replaced by ChatGPT anytime soon. It is not capable of introducing cutting-edge methods that are not already available in the existing literature. In conclusion, while ChatGPT can be a helpful tool, it is important to approach it with realistic expectations and recognize its limitations. Human expertise, critical thinking, and creativity remain vital components of data analysis that cannot be fully replicated by current AI models.

Clarifying that ChatGPT currently has a minor impact on academic research in data analysis, its major influence would be on teaching and general consulting, two seemingly different fields that share a common focus on basic and readily available approaches. The question of whether ChatGPT can replace teachers in the classroom is intriguing and multifaceted, relating to the influence of technology and artificial intelligence on education. To dispel any misunderstandings, technology cannot completely replace the role of a data analysis lecturer in the classroom. Human characteristics such as mentoring, serving as role models, and providing guidance throughout the learning process (beyond just problem-solving) are irreplaceable aspects that ChatGPT cannot emulate. Instead, the lecturer's role in data analysis is to harness technology as a tool, understanding its capabilities and leveraging them to enhance students' own skills. Tools like ChatGPT or future automation can automate routine and repetitive tasks, thereby simplifying the lecturer's responsibilities. ChatGPT not only supports and aids the students' learning process but also contributes significantly to the lecturer's tasks, especially when discussing basic methods and readily available code with minimal effort. However, it remains essential to verify the information provided by ChatGPT, as it may occasionally offer partial or misleading information. Similarly, the lecturer retains the responsibility of defining the learning process, including identifying the problems to be solved, selecting suitable technologies aligned with the learning objectives, and developing strategies for effective use and integration of technology. Nevertheless, there is a challenge in using a tool like ChatGPT, which is still largely unknown in its full potential. Lecturers must periodically update their skills to effectively manage and incorporate technology in the learning process. In this manner, the use of technology can enhance the quality of learning and yield greater benefits for both students and lecturers. Some compare this growing movement to the introduction of calculators in schools, where concerns initially arose about a decline in mathematical skills (which did not occur). Instead, the learning process became more effective with the aid of calculators. Similarly, ChatGPT and other technological tools have the potential to enhance educational outcomes. Overall, while ChatGPT can bring significant advancements to teaching and learning, it cannot fully replace the vital role of human instructors. The integration of technology should be guided by ongoing skill development and careful consideration of its appropriate and effective use in the educational context.

Consultancies may be more significantly impacted by the widespread use of ChatGPT compared to teaching. We are currently witnessing just the tip of the iceberg, as a tsunami of change is likely to follow. Roles in analytics that primarily involve repetitive tasks are expected to be largely automated in the near future, resulting in some job losses. It is crucial for consultants to adapt and seize the opportunity presented by time and efficiency gains. They should focus on developing new skill sets and concentrating on areas where human expertise can truly make a difference. This will be the key objective for consultants moving forward. I view this as a tremendous opportunity, although it also carries significant risks. The generation of routine-based suggestions by ChatGPT may create a perception that anyone can be a data analyst or scientist. However, it is hoped that the risk associated with standardizing methods within consultancies will be counterbalanced by the introduction of more advanced approaches. If the consultancy sector relies solely on basic methodologies, there will be limited room for human involvement in the near future. In summary, while teaching may experience some impact from ChatGPT, the consultancy sector is likely to face greater challenges. It is crucial for consultants to adapt and leverage the efficiency gains offered by automation, focusing on developing new skills and embracing advanced approaches. This presents both opportunities and risks, and it is essential to strike a balance between standardization and the unique value that human consultants can provide.

Antonello Maruotti
Professor of Statistics at the LUMSA University of Rome