Currently, machine learning models are widely used in various professional fields and form the basis of many mobile applications, software and online services. Although many people meet and enter these models, few fully understand their action and basic processes.
In the modern world of machine learning, models are becoming more and more complex and rich in functions. Their increase raises an important question: how can we make these models more understandable and interpretative for a wide audience, including specialists without deep knowledge in the field of machine learning?
Scientists from the University of California, Irvine and Harvard University developed Talktomodel. It is an interactive conversation system designed to explain machine learning models and their forecasts to both professionals and non -experienced users. This interface enables dialogue with ML models using ordinary natural language.
Research is based on previous achievements related to the explanatory artificial intelligence (XAI) and interaction of man-Ai. The main goal of this work was to introduce a new platform, which can provide clear and available explanations of how artificial intelligence works, as is the OpenAI conversation platform, chatgpt, answers questions.
Researchers conducted an experiment with the participation of healthcare employees with different levels of machine science. Almost all participants were new in this field. They were invited to use Talktomodel to answer questions and understand how machine learning models work.
The test results were impressive. Most users preferred to use Talktomodel to understand the models. They completed tasks faster and more accurately using this interface. Even machine learning engineers admitted that Talktomodel is a useful tool.
How does Talktomodel work? Transforms questions into structural logical forms that allow ML models to explain and interpret. This approach ensures flexibility in dialogue, supporting an open inquiry and facilitating understanding of complex models.
Talktomodel is an innovative system that opens the door to natural conversations aimed at understanding machine learning models used in various sets of data and tabular classifiers. Instead of complex programming, users communicate with Taltomodel in natural language (Fig. 1, block 1). The dialogue engine analyzes the input data in the executable performance (Fig. 1, block 2). The performance engine performs operations, and the dialog engine uses the results in its answer (Fig. 1, block 3).
Figure 1. Review Talktomodel
Thanks to Taltomodel, users can discuss why some forecasts appear in the model; changes that will occur in the field of forecasts during the changes in input data; Ways to change forecasts and more. This analysis can be applied to any data group, whether one individual instance or the entire data category.
For example, if you want to predict the development of diseases, you can ask questions such as: “How important is the body mass indicator (BMI) for forecasts?” or “How is the probability of changing the disease after lowering the glucose level by 10 in men over 20 years old?” Talktomodel will provide information by saying that BMI is the most important predictive feature, and a decrease in glucose levels by 10 will reduce your chance of developing diabetes by 20%. Then you can continue the dialogue by asking additional questions. Talktomodel makes it easier to explain how models work because you can talk to the system in natural language and give you an informative answer.
You can see an example of such a dialogue in Fig. 2.
Dig. 2
To support significant conversations with Talktomodel, there are methods of improving language understanding and modeling. First of all, the dialog engine was implemented, which analyzes the introduction of the user's text. These data are converted into a language similar to a structural language language using a large language model (LLM). LLM performs analysis by treating the task of translating users 'statements into the programming language as a problem of learning SEQ2SEQ, with users' statements as a source and analyzing the programming language as a goal.
In addition, the TalkTomodel system combines explanation operations, machine learning errors analysis, data manipulation and described text generation in a single language, which may include a wide range of potential conversation topics that are needed in the most explaining models. Examples of various operations are presented in Fig. 3.
Figure 3. Operations are included in the conversation to generate answers.
The system offers an operating mechanism that automatically selects the most appropriate explanations and operations for the user. This reduces the load on users and makes the interaction with the machine learning model more available. In addition, a text interface was created, which even allows people without high technical skills to understand and interact with ML models. As a result, Talktomodel means that explaining how machine learning models work more accessible and understandable to a wider audience.
In the future, the use of Talktomodel may expand to use the system in real clinical and laboratory conditions in which participants can use it to understand and optimize models. In addition, future research can focus on visualization and analysis of raw data to increase user trust.
Talktomodel is a step forward in the development of the field of explanation of artificial intelligence. This interface allows you to talk to complex machine learning models in natural language and understand their decisions. This tool promises that ML will be more accessible and interpretative for everyone.
You can find the model code Girub.
