Expertise and contributing factors

Expertise is at the core of Starmind’s AI. It is Starmind’s internal representation of the skills and knowledge of each user.

A user’s expertise on any topic can grow or decrease depending on how the user interacts with this topic on the Starmind Q&A application or on other connected applications (see "Positive factors" and "Negative factors" below). The expertise can also change indirectly over time or through the interaction of other users with the same topic (see "Forgetting" below).

Starmind’s expert search algorithm uses expertise to determine who is an expert on a given topic. An expert search is triggered, for example:

• When publishing a new question to determine who should be notified about this question.
• When searching for experts on specific topics on the Experts page in Starmind’s user interface.

Expertise is also used to select the most relevant topics to be shown on the profile of each user. When a user wants to manually add additional topics to their profile, they are included immediately if the user already has expertise on this topic. Otherwise, the topic will be pending until the user obtains expertise on this topic according to Starmind’s AI.

Positive factors

The following factors lead to an increase in the expertise of a user. When the expertise of a user on a given topic grows, then it becomes more likely that this user will be listed as an expert on this topic and more likely that this topic will appear on the profile of the user.

Data from connected external sources

Starmind can connect to a variety of applications and platforms such as Microsoft Teams, Slack, calendars and document management systems like SharePoint or OneDrive. This allows Starmind to learn about the expertise of users directly from the different applications they use in their daily work. Each connector allows Starmind to learn from different kinds of data. Details can be found on the Connectors page.

Answering questions

A user’s expertise on a topic grows by answering a question on the topic. The answer must either still be unrated or have an average rating of at least 2.5 stars. Better ratings lead to a bigger increase in the user’s expertise.

Expert recommendations

A user’s expertise on a topic grows by being recommended by another user as an expert for a question on the topic.

Asking questions

Asking a question on a topic leads to a weaker increase in expertise compared to answering or being recommended as an expert. If the user’s expertise on a topic comes exclusively from asking questions, then the user will not be shown as an expert on this topic. This protects the anonymity of the question poser, especially for questions on new topics.

Job titles

When Starmind recognizes the job title of a user, then Starmind will assign expertise to the user on the skills that are required for this job or commonly found among other users with the same job title.

The associated expertise is relatively weak. The common job skills primarily help to build an initial profile for new users. Once Starmind learns more specific topics of expertise for a user, the expertise on these specific topics soon becomes stronger than the expertise that was only based on the job title.

Job titles can be provided for all users in the organization through SSO claims. Alternatively, each user can enter their job title individually in the Starmind application.

To optimize how Starmind learns expertise from job titles, please consider our recommendations on job titles.

Negative factors

The following factors cause the expertise of a user on some topics to decrease.

Badly rated answers

When an answer gets an average rating of less than 2.5 stars, the expertise of the user who wrote the answer is reduced for the relevant topics.

Rejecting expertise

The Starmind application gives users who have been recommended as an expert for a question the opportunity to indicate that they are not able to answer the question. This applies both to experts that were selected by Starmind’s AI as well as to experts that were recommended by other users. If the recommended expert gives such feedback, then their expertise for the topics of the question is reduced.

Forgetting

Knowing how to forget is an important ability of Starmind’s learning algorithm. Forgetting ensures the expertise of each user is always up-to-date, even when none of the factors listed above occur.

Forgetting induced by other users
Every time a user gains some expertise on a topic, the expertise of all other users on this topic is slightly reduced. This ensures that new users always have a chance to quickly establish themselves as experts on a topic, replacing pre-existing experts who no longer contribute to this topic.

Forgetting over time
Expertise slowly diminishes with time. This ensures that users who no longer contribute to a topic will gradually lose their expertise, even if no new experts take their place.

Learning about interests of users

Starmind uses additional factors to learn about the interests of users. Such factors don’t affect expertise, as a user might be interested in a topic without being an expert on that topic. For example, Starmind remembers the topics of questions and answers each user is often searching for. Such topics will not show up on a user’s profile (except for manually selected ones) and the user will not be listed when searching for experts on these topics. Instead, these topics are used for example to give personal recommendations to users about what questions and answers are relevant for them.