Bad Incentives: How the ServiceNow Knowledge Base Resembles the Prisoners’ Dilemma, Part 1
ServiceNow Knowledge Base Background
ServiceNow is a Platform as a Service (PaaS) that offers an array of Service Management applications, with roots in IT Service Management (ITSM). Among them is a knowledge base, where users can gather, analyze, store, and share knowledge with the purpose of improving efficiency by reducing the need to rediscover knowledge. 
Prisoners’ Dilemma Background
The prisoners’ dilemma is a classic thought experiment that shows an example of where people might not cooperate, even though it is in their best interest as a group to cooperate. In the classic example, there are two prisoners who are being interrogated for a crime. Each prisoner has two choices – either turn in their partner for a reduced sentence (betray), or stay silent (comply). For this example, we’ll say that if both prisoners comply, and stay silent, then they both will only get a 1 year sentence. If they both betray each other, then they’ll each get 3 years. But if only one prisoner betrays the other, then the prisoner who betrays will be released, while the prisoner who stayed silent will get 5 years. Each prisoner has to make the choice independently and without discussing with each other– they can’t collude to both stay silent. The chart below summarizes the options and payoffs:
Clearly, the prisoners would be better off as a group if they could both stay silent – there would only be 2 total years of prison time. However, this scenario tends to end with both prisoners betraying. A double betrayal is what’s referred to as the Nash Equilibrium, which essentially means that this is the most likely outcome (see the link for more information).
Implications for ServiceNow Knowledge
Fortunately, the stakes are much lower for ServiceNow knowledge bases. No one is going to prison (at least, I certainly hope not!) However, the basic structure of the prisoner’s dilemma can potentially apply to the Knowledge base. Using the prisoners’ dilemma as a framework, I’ll discuss some aspects that can lead to a lower quality knowledge base and specific examples of how to fix these issues.
Imagine that instead of prisoners, the two parties are the authors of articles and the readers of the articles. The parties have two choices, similar to the prisoners’ - they can either “cooperate”, or “betray,” meaning that they can either do something beneficial to the knowledge base, or something more selfish. For the author, this would be either writing a high quality article that includes some of the attributes previously described, or writing a sloppy article. For the reader, the choices would be either checking the knowledge base for the answer, or just emailing the author for the answer. Rather than prison time, the outcomes are measured in the total amount of time that the person has to spend for his portion of the knowledge base. Again, in this thought experiment, there are no outside agents, and the author and readers make their decisions in a vacuum.
How is This Similar to The Prisoner’s Dilemma?
Though the numbers are different, this thought experiment is structured in such a way that the Nash equilibrium is the same as before – with both parties betraying each other.
Why does this happen? There are several problems that lead to a Nash Equilibrium outcome of a double betrayal: 1) there’s an incentive structure that discourages cooperation, 2) the parties are only interested in themselves, and 3) there’s a lack of trust between them.
We’ll discuss reasons 2 and 3 in future articles, but for now let’s examine reason 1 – the incentive structure – from the prisoner’s dilemma and see how it applies to this Knowledge situation.
Misaligned Incentive Structure (From Prisoner’s POV)
If the warden was attempting to create an incentive structure that would encourage the prisoners to betray each other, he got it.
The reason for this is that they will individually get less jail time by betraying, regardless of what the other prisoner does. For example, if prisoner A stays silent, prisoner B will go free by betraying, but would get 1 year by staying silent. Or if prisoner A betrays, then prisoner B only gets 3 years by also betraying, versus 5 years by staying silent. In both scenarios, prisoner B is better off by betraying, and the same thing happens for prisoner A. This leads to the Nash Equilibrium outcome of both prisoners betraying.
Misaligned Incentive Structure (from ServiceNow POV)
The incentive structure in the ServiceNow Knowledge base is similar to the prisoners’ dilemma – the optimal solution to minimize total time involved is for everyone to cooperate and do the best job they can (write a good article, and check the knowledge base for answers).
Similar to the prisoners’ dilemma, the author has the incentive to write a sloppy article because regardless of what the readers do, he will spend less total time if he writes a bad article (if the readers check the knowledge base, he would spend 1 hour total on a good article, versus 50 minutes total on a sloppy article). Similarly, the readers have an incentive to just email the author, since they will end up spending less time total by doing so, regardless of the quality of the article that the author writes (1 hour and 20 minutes on the good article side, and only 50 total minutes in the bad article case). Assuming that the time estimates are reasonable, the incentive structure leads to the same Nash Equilibrium from before – with both parties betraying. Regardless of the specific time estimates of this example, it’s easy to imagine a scenario where both parties would have an incentive to slack.
How Can Knowledge Admins Avoid Bad Incentive Structures?
In the classic prisoners’ dilemma, the incentive structure is set up in such a way that can incentivize users to act in a way that’s detrimental to the overall health of the base. It’s important for Knowledge Administrators to recognize this pitfall, and ensure that appropriate steps are taken to remove these temptations. We’ll discuss two ways to fix this bad incentive structure – altering the incentive structure, and changing users’ perceptions of success.
1. Change the Incentive Structure
This might seem obvious, but the best way to prevent users from falling into this pitfall is to change the incentive structure to reward the behavior that makes the base work properly. You can do this by rewarding good efforts, and making sure that poor efforts have to be corrected.
Reward Good Efforts
Make sure that authors and readers are properly rewarded when they do a good job. These rewards need to be on-going to make sure users continue to perform their duties effectively.
Rewarding good efforts can help fix the misaligned incentive structure. In the example above, the only rewards are in terms of time spent looking for the solution, or time spent writing an article. Additional incentives can be created to help motivate users to maintain a high quality knowledge base.
An example that a knowledge admin could implement would be recognizing authors that consistently produce high quality articles during yearly reviews. Quarterly rewards (giftcards, etc.) given to the authors with the highest rated articles would also help encourage good content.
Implement Systems to Improve Poor Content
This could be thought of as the opposite of “rewarding good efforts.” If you force an author of a sloppy article to rewrite the article, you are essentially punishing the sloppy writing habits. This will continue to fix the misaligned incentive structure – if the author has to spend extra time fixing a bad article, then maybe it would just be quicker to write it well in the first place.
How to Improve Poor Content
As far as improving poor content, CapTech suggested a number of strategies to a Fortune 500 client. Stay tuned for a future blog post about improving low quality knowledge articles!
2. Change Users’ Perceptions of the Game
The other way to change the incentive structures is to change authors’ perceptions of the incentives. The assumption we’re using above is that users are only thinking about how to minimize the time that they individually spend in the knowledge base. But what if you can change the users’ perceptions, and have them not be only focused on their own time, but of the total time spent in the knowledge base. This can make a huge change to the fundamental incentive structure. We will discuss how to make this shift in Part 2 of this series, which will be released soon.
Don’t Like the Outcome? Change the Game.
Throughout history, entrepreneurs have succeeded by changing the rules. When it comes to ServiceNow Knowledge bases, the rules of the game can be stacked against you. As a Knowledge Administrator, it’s important to recognize that the incentives in the ServiceNow knowledge base may be stacked against you. But you have the power to change the rules, and alter the incentives to encourage people to use the Knowledge base in a productive manner.
In the next post in this series, we’ll discuss how to shift users’ perceptions of the incentives of knowledge – to get users to think of the entirety of the knowledge base, not just themselves.