Ai Incubator
An Ai model to improve and monitor financial regulation and compliance
Problem Statement
in progress
Objective
in progress
User
in progress
Confidential Client Team
in progress
Methods
Requirements & Constraints
Task Analysis
Design Review
Prototyping Feedback & Testing
Usability Testing
Tools
Analog Sketching
Figma
Team
Abhishek Kulkarni
Amia Oberai
Madelyn Sando
Michael Seals
Process
in progress
Design Iterations
in progress
The intelligent recommender system determines violation dot placement. Based on the user, what they care about, and their role in the company, the system will place violation dots in proximity to one another based on how similar they are (based on the criteria that matters to the user), Similar to the aforementioned Internet Map. Dots or dot clusters the system considers to be most relevant to the user will be placed closer to the center of the map, while those considered less relevant to the user will be placed towards the periphery of the map. By default, the dot map will start out zoomed in to the center (shown by the box with the dashed lines), thereby showing only the violation dots the intelligent recommender system considers most relevant to the user. The user can zoom out to see other dots as well; if they zoom out all the way, they will be able to see every violation in the database! Prototype - Full View Great. that you so much Abhishek. Now I am going to walk you through the actual prototype and so you can see how we envision these concepts and inspirations working together. To review the general layout of the interface - First you can see that we have several sections of the screen. The two largest sections are a data cloud point on the left (for you to examine individual or clusters of data) and a newsfeed with relevant violation information on the right. The newsfeed will use a recommender system to order the content based on the users needs - irrespective of if anything is selected or not You will also see the top bar offers search and filtering navigation. While in this view we see a large portion of a data cloud, in reality a user will only see the data points that are in the dotted box. And this is a feature of the recommender system that we are proposing. While there are many data points within the data cloud, the recommender system will help the user focus on what it thinks would be the most appropriate cluster by offering them the zoomed in view that you see on the next slide.
When the user double clicks on a violation in the dot map, more details such as the violation summary will be shown in a popover for the user to reference. This ensures that the user can select a dot without being hit with a distracting popover, but can opt to see one for progressive disclosure if they so choose. Prototype - Single Dot Selected The bottom informational panel will update to display information about whatever the specific violation the user has selected. It will display information such as the company and the fine amount. The news feed on the right will similarly update to show stories pertaining to the specific dot selected. For instance, if the dot selected is about a fine paid by the Intercontinental Group, news stories about that violation will populate the news feed.
Prototypes - newsfeed Detailed view The news feed uses an intelligent recommender system to curate and order content for the specific user based on their history and their role in the company using natural language processing to understand news articles and both content-based and collaborative-based filtering to curate the articles for the user. The content is dynamically updated based on whether and which violations are selected in the dot map to remain relevant to the user’s focus and task. If the user selects a news story, all of the dots in the dot graph related to that story will automatically be selected. The user may like or comment on stories in the news feed. These likes and comments will be internally visible to other users of our platform, but not to the outside world. This is intended to support communication amongst policymakers across the organization, as the client had requested. Likes and interactions also offer a powerful way to inform the intelligent recommender system on which articles are currently the most important and relevant. If the user feels a given news story is not relevant to them, they can hit the X button on the story’s card to let the intelligent recommender system know so it can improve its recommendations in the future (similar to facebook). Next we look at a a zoomed in view of the previous point cloud. And you can see that we have a bottom navigation bar at the that provides more description regarding the data points that you have zoomed in on. Simultaneously you will also notice that the newsfeed has updated to only include data that is relevant to the viewable area of the data cloud (similar to how google maps updates their offerings based on your view)
Dot Cloud Legend: explanation of interactivity and variables This large web-map format of the violation shows the connections between different violations. This map has interactivity that is similar to google maps Each data point in this dot cloud represents individual violations that can be categorized and connected based on a few indicators that are unique to each piece of data: Size of Circle - size of fine Color of Circle - type of violation Saturation of Circle - newness of violation (older violation appear to fade away) Options to Sort And we take you through some of the features in more detail-
Prototype - 2 Dots Selected When multiple dots are selected, the system will intelligently determine what all of the dots selected have in common (same company, same violation type, etc.) The bottom informational panel will update to display information about whatever the selected dots seem to have in common (so, for instance, if all of the dots are about money laundering, the panel will display information about money laundering. If all of the dots related to the same company, the bottom panel will display information about that company). The news feed on the right will similarly update to show stories pertaining to what all of the dots have in common. For instance, if all of the dots pertain to the Money Laundering, news stories about Money Laundering will populate the news feed.
Prototypes - Search/Filter Detailed view Explain features/options And now we can examine the search bar at the top, if for example you wanted to search money laundering. you can see that the dot cloud would automatically update to only show the relevant data points. As you can see in the filtering options, users would be able to filter based on type of violation, rule, fine amount, company name, or date. These options are just some parameters that we thought were important for refining a data search and can be customized to more appropriate filters for a team. As you can see a drag option also would allow users to see know many options are available at a given range (much like the dynamic pricing updates you would see on airbnb/zilow). The user cna manipulate this to further refine the search. And finally a user can save a search for future viewing. We imagine this scenario would be appropriate when a user has specific preferences that they would like to view again later. And next Mike is going to walk you through the newsfeed in more detail
Results & Reflections
in progress
Next Steps
in progress