Using Ranking to Reflect Intelligence in User Interfaces
From search engine results to e-commerce recommendations, ranking patterns can be found practically everywhere in digital products. Ranking is such a pervasive pattern because it helps users easily identify quickly the top options. In fact, using ranking in user interfaces (UIs) helps business users find the best options from a given dataset.
There are many scenarios where ranking is useful in enterprise software. In a procurement scenario, business users need to understand which is the best provider to procure a good or service. Another example can be in professional services, where project managers want to understand which is the best talent from an available group of consultants.
From a user perspective, ranking helps to order a group of elements with a common comparable value to determine the best option, preference, value, relevance, or priority.
Ranking and intelligent systems
Ranking is among a group of basic patterns that can help to reflect an intelligent insight. Ranking, matching, comparing, feedback, and intelligence explanations provide affordances and additional interaction points that business users can use to work with artificial intelligence (AI) output. In an enterprise software setting, there are plenty of examples where a ranking pattern can support intelligence scenarios. Here are some examples:
One of the areas where an intelligent system can support humans is on multiple-criteria decision making. An intelligent system can quickly go through multiple dimensions and provide a ranked group of items for further analysis and evaluation. A much-discussed approach for this would be in the health care sector, where intelligent systems could support experts to improve early cancer diagnosis.
Automation is a prominent use case for AI. Think of the growing capabilities of self-driving cars. For enterprise software, one possible use case could be related to asset management, where a system uses computer vision to evaluate possible damage to a fleet of trucks, generating a list of service orders ranked by a particular severity or importance.
One use case for this kind of scenario could be an intelligent system using generative design to assist a designer or engineer to go through hundreds of variations on a design that otherwise would not be humanly possible to generate. The system could provide a ranked iteration list based for example on cost reduction. Business users might use this to analyze, further explore, and tweak any proposals.
Providing suggestions based on behavior patterns is already heavily used in e-commerce websites. As I mentioned already, in a procurement scenario, business users might need to understand which is the best provider option to procure a good or service. Instead of providing a full list of providers, an intelligent system could suggest the top five providers that would fit a particular order. This could make the procurement process nimbler and more straight-forward.
Predictive analytics is a good example of this category. For example, Artificial intelligence and machine learning can help leverage the information available to a CFO to better forecast things like company revenue or risk. Business users can be presented with a list of possible scenarios ranked by feasibility.
Considerations for using ranking in UIs
Ranking and sorting are two different things. But because ranking already denotes an intent of order, a ranking will be coupled with a sorting (sort by rank) when reflected in a list or table.
Presenting a ranked data-set in a table does not take away the usual attributes like sorting, filtering, column order, search and grouping. For example, end users should be able to sort any particular table in a different way to the rank order initially presented. An example would be providing a business user with a table ranked by a score, and the business user changing the sorting by price to accommodate a particular task he or she wants to perform.
If you add semantic colors, provide ranking tags (e.g. Good, Best Option, Average), or group the ranked elements, you will create thresholds at the boundaries of these groups.
An important scenario to consider is when very similar items end ups on different sides of the thresholds. This might prompt the end user to think why, for example, Provider A is labeled as good and Provider B as average, even when they are very similar in the eyes of the user. Here is where an explanation comes in handy to clarify and provide the user with context.
If we aim to augment and support our human workforce with the use of AI, we need to provide the transparency required for business users to accept intelligent system recommendations.
Providing an explanation on how the rank was put together will allow end users to better understand the information with which they are presented. This will not only contribute to building confidence in the system, but will also allow business users to have enough understanding to provide feedback and judge the quality of the AI output.
This is just the beginning of the journey
These are some of the thoughts we have encountered while working on ranking for intelligent systems. It is not a complete list by any means. We will certainly find more aspects to consider as applications start embedding intelligence in their systems and our business users start consuming them.