Using AI Tools for Research
Case studies
-
Data exploration:
-
Accessing and utilizing high-quality datasets is critical in research. AI can explore a dataset and streamline data cleaning processes by automating the identification of duplicates, inconsistencies, and errors in large datasets. AI can substantially speed up the data cleaning and preparation phase—tasks that often take a significant amount of time. By automating the detection of duplicates, inconsistencies, missing values, and outliers, researchers can focus more on analysis rather than preparation.
-
For example: You are a Ornithologist that has acquired a large dataset from various citizen science websites, containing the past two decades of bird sighting data. The datasets you got are too large to examine one by one. In addition, you notice some missing values, errors, as well as duplicates in the subset you were testing. You turn to the Hopkins AI Lab to help you explore these datasets and come up with some strategy to process, clean, and analyze these large datasets.
-
Caveat: It is your responsibility to ensure that the suggestions provided by an AI tool are logical and accurate. Never upload a data file to an AI tool, ask it to process and analyze the data, and blindly rely on the results without understanding how the data was processed or analyzed.
-
-
Apply AI in GIS
-
AI tools in ArcGIS, or GeoAI, can be used for feature extraction, image analysis, and predictive modeling. Esri has a growing library of Pre-trained AI models that are designed to extract features from different data sources with minimal setup. These current models are available for common workflows. Users can also train their own custom models. Esri has also added AI assistants to a number of applications such as Survey123 and ArcGIS Business Analyst.
For example:
-
You are glad that Spring is finally here because you live in Baltimore City, and you know that workers will start filling in the many potholes that has caused damage to your car. You heard that city workers use the Pavement Crack Detection Pre-Trained Model to map where the worst roads are located so the repair work can start as soon as the weather cooperates.
-
You are trying to determine the devastation after a hurricane. You have satellite imagery of the before and after in your area. Within ArcGIS, you can use the Building Change Detection Pre-Trained Model to locate the buildings that incurred damage or was completely washed away.
-
You want to collect data about the varying public perceptions of safety in your city. You realize having the various neighborhood groups spread a survey would be much more efficient than going door to door. Since location is important, you know you want to use ArcGIS Online’s Survey 123 tool, but you have never designed a survey before. Once signed, you can choose to use the Survey 123 Assistant to help you, then modify it until it is ready for publishing.
-
-
-
Code debugging
-
AI-powered debugging tools leverage machine learning algorithms to analyze code patterns and quickly identify syntax errors, logical errors, and performance bottlenecks. These tools can often provide suggestions for fixes or improvements, making them invaluable for researchers, coders, and students.
-
For example: As a researcher that is developing a machine learning algorithm to analyze MRI scans for early detection of brain tumors, you encounter bugs within your Python code, which is leading to inaccurate model predictions. Because traditional debugging methods consume a significant amount of time, you use the Hopkins AI Lab to help you decipher the error messages and provide suggestions to resolve the issues more quickly.
-