The Effect of Transparency of Algorithm on Decision-making

---

Research shows that evidence-based algorithms have consistently outperformed human professionals when making predictions. But people still don’t want to use algorithms to make decisions --- known as algorithm aversion.  

The research I conducted focuses on examining: 

- Why do people show algorithm aversion?  

- The impact of transparency of algorithm on people’s willingness to use the algorithm to make decisions?   

Participants: 

685 participants recruited from AmazonMechanical Turk (Mturk) 

Method: 

Survey questionnaire using Qualtrics  

- People who received a detailed description of how algorithms worked were more likely to use algorithms compared to people who received a lay-person description in a hypothetical scenario ( hiring). 

- Prior knowledge about the context of using algorithm matters (detected significance in hiring situation but not in more technical situation, predicting the rank of 10 US states in terms of number of airline passengers who departed from that state in 2011).

- If we want people to use algorithms to make decision, provide them with more information about how algorithm is built. 

-  Significantly improved high-stakes decision-making for companies in multiple domains.