Learning & Innovation Research Manager| Project Management Institute (PMI)Spain
Data quality and quantity is particularly important as we think about leveraging AI on projects. Considerations include the diversity and comprehensiveness of the data that is available to us.
Have you ever encountered unexpected challenges or pitfalls while using data in your projects? How did you navigate the situation and find a resolution? Saving Changes...
When it comes to unexpected data challenges, I start by getting curious and look at how and where the data is being pulled. I want to know the data is reliable, so I'll test by pulling the data in different ways, comparing sets, etc. to make sure the data is "clean." I'll also pull in colleagues to make sure I'm not missing something. Saving Changes...
It's recommended to scrutinize the data before incorporating it into the model. However, when encountering unexpected data that could influence the outcome, it's prudent to refine your query to identify which data should be considered for a specific task. Additionally, categorizing your data and filtering out irrelevant information will ensure the model delivers the highest quality output.
This makes sense and applicable. Saving Changes...
Engage and understand the threat, assess risk and ensure that we're monitoring the trheat once identified while planning our response. Saving Changes...
Definitely, I’ve run into issues with data quality before. Once, we faced problems with incomplete and inconsistent data, which threw off our analysis. We tackled it by first figuring out exactly where the gaps and inconsistencies were. Then, we cleaned up the data by filling in missing values and standardizing entries. It took several rounds of cleaning and checks to get it right. We also cross-checked our data to ensure accuracy. It was a bit of a process, but it was crucial for making sure our AI models performed well. This experience really underscored how important good data preparation is.
I am managing a project that was abandoned after a serious data breach. I feel that it lacked a basic understanding of data management. So I am ensuring that there is a proper data management policy incorporated and well-defined processes as well as protocols for certain situations. The idea is to ensure the anticipated situations and build solutions considering the situations. So even if there is a breach, safety mechanisms like siloing data and anonymizing or encryptions or tiered user roles with need-to-know basis access policy are being used to ensure that any unexpected challenges could have minimum impact on the project. Saving Changes...
Key to understand is what you stated: "Data quality and quantity is particularly important as we think about leveraging AI on projects". This is key in AI from long time ago, from 1970s at least. But all related to data rest on other discipline that today is called Big Data to put it under and umbrella. And it is independent of you use AI or not. So, let me say, nothing new below the sun. Just to understand that data has to be converted into information. Again, nothing new. It was analized by Claude Shannon in the 1940s.
however Sergio even that many tools can be applied specialli to AI data, video manipulation, fake news can give you strange allucinations so that human control at the end is the only way that AI unexpected outcomes can be monitered, detected and corrected .... Saving Changes...
I have encountered unexpected challenges when using data from electrical system variables, specifically frequency and voltage. A frequent issue was the inconsistent quality and lack of integrity in the collected data, which made it difficult to draw valid conclusions. To address this, we performed extensive data cleaning, which included identifying and removing outliers and measurement errors, imputing missing values, and normalizing the data. These actions resulted in a clean and reliable dataset for frequency and voltage variables, leading to more accurate and useful analyses. Saving Changes...
How do you navigate unexpected data challenges in your projects?
It comes in many forms in my short experience. Whether that being the integrity of the data, loss of the data and other instances. When I have had issues with corrupted data it has varied from just going back to the original source and seeing if there was a tech error and in some cases having to start over from an organization's backup. It can be time consuming but worth it for the client. Having to scrap data is no fun for anyone. Saving Changes...
Yes, when it is evident that the data collection was not correct, obtaining non-standardized data. In that case, it was necessary to clean the information. Saving Changes...
PMI Team Member
