by Adam Markiewicz
The context
I’ve been interested in data science since early 2003 when I used a few statistical methods in my master thesis on hedging electricity price risk. However, my work had not exposed me to any substantial data challenges until around 2015 when I became involved in data-heavy contentious matters, including a few competition-related disputes requiring the use of advanced econometric modelling.
Still, until now, MS Excel remained by far the number one data-processing tool me and my colleagues used to deliver forensic expert assignments. Increasingly, I became aware of Excel’s inability to efficiently handle large datasets and complex data modelling and realised how many of my past engagements could have been delivered more efficiently with other data science tools. This realisation led me to complete various training on the topic, including an excellent Applied Data Science program by MIT Professional Education.
Alongside the recent discussions on the benefits and controversies of ChatGPT, I’ve decided to kill two birds with one stone and test both ChatGPT’s conversational ability on a professional subject and get its ‘understanding’ of the forensic accounting profession and its recommendations on the use of machine learning in the context of expert evidence on quantification of damages.
Here is the transcript of my conversation conducted with ChatGPT Feb 13 Version on 13 March 2023, just before the most recent ChatGPT 4 was released. Some fragments of the chat have been redacted for the sake of brevity, but the transcript is still long. If it’s too long for you to read it all do scroll to the end for my ‘takeaways’ and conclusions.
The interview by Adam
Hello! Can you introduce yourself, please?
ChatGPT: Hello! I am ChatGPT, a conversational artificial intelligence language model developed by OpenAI. I have been trained on a vast amount of data to understand and generate human-like responses to a wide range of topics. My goal is to help people by providing information, answering questions, and engaging in conversations. […]
OK. Can you explain, what does a forensic accountant do?
ChatGPT: […] A forensic accountant is a specialised type of accountant who uses their skills and knowledge to investigate financial crimes or disputes. Their work often involves analysing financial data and records to uncover any irregularities or inconsistencies, and they may be called upon to testify in court as an expert witness.
Forensic accountants can be involved in a wide range of cases, such as fraud investigations, insurance claims, divorce settlements, bankruptcy proceedings, and more. They may work for government agencies, law firms, accounting firms, or as independent consultants. […]
Thanks! What is exactly the role of an expert witness?
ChatGPT: An expert witness is a professional who is qualified to provide expert testimony in a legal case based on their specialised knowledge, training, and experience in a particular field. The role of an expert witness is to provide the court with objective and impartial opinions on technical or complex issues that are relevant to the case.
The expert witness may be called upon by either the plaintiff or the defendant, or they may be appointed by the court to provide an independent opinion. Once they have been called upon to provide expert testimony, the expert witness will be expected to provide a clear and concise explanation of their opinions to the court.
The opinions of an expert witness can be used to help the judge or jury better understand technical or complex issues that are relevant to the case. […]
Overall, the role of an expert witness is to provide objective and impartial opinions based on their specialised knowledge, training, and experience, to help the court make informed decisions about the case.
OK. How can you potentially help a forensic accountant to perform better as an expert witness?
ChatGPT: As an AI language model, I can potentially help a forensic accountant to perform better as an expert witness by providing them with relevant information and resources to support their testimony. Here are some ways that I can be of assistance:
Providing research: I can help the forensic accountant to find relevant research studies, cases, and other materials that support their opinions. This can help to strengthen their testimony and provide a more robust foundation for their arguments.
Answering questions: The forensic accountant can ask me any questions they have related to the case, and I can provide answers based on my training and knowledge. This can help them to better understand technical concepts or to clarify their understanding of complex financial data.
Practice sessions: I can simulate a courtroom environment and provide feedback on the forensic accountant's performance as an expert witness. This can help them to improve their communication skills, to develop clear and concise explanations of complex issues, and to prepare for challenging questions that they may face during cross-examination.
Providing examples: I can provide examples of other cases where similar issues have arisen and share how other forensic accountants have successfully navigated these challenges in their testimony. […]
Can you help an expert witness to be more impartial?
ChatGPT: Yes, as an AI language model, I can provide guidance on how an expert witness can maintain impartiality and objectivity in their testimony. Here are some tips:
Stick to the facts […]
Avoid conflicts of interest […]
Be transparent […]
Acknowledge limitations […]
Be prepared for cross-examination […]
I understand. And how can AI, or machine learning, help a forensic accountant as an expert witness to deliver quality work products given the increasing data intensity in disputed matters?
ChatGPT: AI and machine learning can be powerful tools to help forensic accountants as expert witnesses deliver quality work products, especially in cases involving large amounts of data. Here are some ways AI and machine learning can be used to support forensic accountants:
Data analysis: AI and machine learning can be used to analyse large volumes of financial data quickly and accurately, identifying patterns, anomalies, and other insights that may not be immediately apparent to human analysts. […]
Fraud detection: AI and machine learning algorithms can be trained to recognise patterns of behaviour that are associated with fraudulent activity. […]
Predictive modelling: AI and machine learning can be used to build predictive models that can help forensic accountants to forecast future trends and identify potential risks. […]
Document review: AI and machine learning can be used to review large volumes of financial documents quickly and accurately, identifying key information and flagging potential issues for further investigation. […]
Case management: […]
Can you be more specific, please? Can you give a few examples of machine learning models and their specific applications in, say, quantification of damages?
ChatGPT: Certainly, here are a few examples of machine learning models that can be used in the quantification of damages in legal cases:
Regression analysis: Regression analysis is a machine learning model that can be used to identify the relationship between two or more variables. In the context of damages quantification, regression analysis can be used to identify the factors that contributed to a loss of revenue or profits, such as changes in market conditions or consumer behaviour. […]
Decision trees: Decision trees are a machine learning model that can be used to identify patterns in complex data sets. In the context of damages quantification, decision trees can be used to identify the factors that are most likely to contribute to a loss of revenue or profits, such as changes in pricing or product availability. […]
Neural networks: Neural networks are a machine learning model that can be used to identify patterns in large, complex data sets. In the context of damages quantification, neural networks can be used to identify the factors that are most likely to contribute to a loss of revenue or profits, such as changes in consumer behaviour or market conditions. […]
Random forests: Random forests are a machine learning model that can be used to identify patterns in large, complex data sets. In the context of damages quantification, random forests can be used to identify the factors that are most likely to contribute to a loss of revenue or profits, such as changes in pricing or product availability. […]
Pretty impressive. How can such machine learning models be implemented by a forensic accountant?
ChatGPT: Implementing machine learning models in the context of forensic accounting requires several steps:
Data collection and cleaning […]
Model development […]
Model deployment […]
Model interpretation […]
Overall, implementing machine learning models in the context of forensic accounting requires a combination of technical expertise, data analysis skills, and a deep understanding of the legal and regulatory frameworks that govern financial reporting and disputes. Forensic accountants may need to work with data scientists, software developers, and attorneys to ensure that the models are developed and deployed in a way that is useful and appropriate for the legal case at hand.
What specific tools and software would you recommend a forensic accountant to use to develop machine learning model for use in disputes?
ChatGPT: There are a number of software tools that forensic accountants can use to develop and implement machine learning models for use in disputes. Here are some examples:
Python: Python is a popular programming language that is widely used in data science and machine learning. There are many libraries and packages available for Python that are specifically designed for data analysis and machine learning, such as scikit-learn, pandas, and NumPy.
R: R is another programming language that is widely used in data science and machine learning. It has many built-in statistical and graphical functions that are useful for analysing data and developing machine learning models.
Tableau: Tableau is a data visualisation tool that can be used to create interactive visualisations and dashboards. It can be used to analyse data and present the results of machine learning models in a way that is easy to understand.
MATLAB: MATLAB is a programming language and environment that is widely used in scientific computing and engineering. It has many built-in functions for data analysis and machine learning and is particularly useful for developing and testing algorithms.
RapidMiner: RapidMiner is a data mining and machine learning platform that allows users to build and test predictive models without writing code. It includes a visual interface that makes it easy to build and evaluate models and supports a wide range of data sources and formats.
These are just a few examples of the many software tools available for forensic accountants to use in developing and implementing machine learning models for use in disputes. The choice of tool will depend on the specific needs of the project and the expertise of the forensic accountant.
Very informative. Thank you for your suggestions and this conversation.
ChatGPT: You're welcome! It was my pleasure to assist you. Don't hesitate to reach out if you have any more questions in the future.
My conclusions
I was pretty impressed in ChatGPT’s ability to explain somewhat specialised topics in layman’s terms, which I found very useful to improve the general understanding of forensic accounting. The conversation also looks promising in terms of potential help forensic accounting experts may get from AI in explaining complex concepts and analysis to a non-specialised audience such as the legal community.
I found no errors nor any controversies in ChatGPT’s answers although they seemed a bit general. Its ability to engage in more detailed and specialised conversations remains to be tested, especially in light of ChatGPT’s ‘own’ promise of it being able to assist in expert witness preparations for cross-examination during hearings.
Finally, I certainly do agree with ChatGPT’s suggestions as to the models and tools recommended for use by forensic accounting experts. I expect to see more of them employed on actual cases both by Quantuma and the larger forensic community.
About the team
Adam is a Managing Director in our Forensic Accounting & Investigations team. Our team of experts are used to securing, analysing and presenting data in a forensically sound matter, such that it can be used in evidence or in support of an investigation. The team has worked on a wide variety of matters involving millions of documents and incorporating newer analytical processes such as Machine Learning and we have extremely sophisticated AI capabilities within our eDisclosure review platform.
