Resolving the Limitations of the CNI Model in Moral Decision Making Using the CAN Algorithm: A Methodological Contrast.

The CNI model generates C , N , and I parameters to measure people's mental processes-consequence sensitivity ( C ), norm sensitivity ( N ), and generalized inaction/action preferences ( I )-in moral decision making. Given the limitations of the CNI model, the CAN algorithm was developed to depict the consequence sensitivity ( C ), overall action versus inaction preferences ( A ), norm sensitivity ( N ), and perverse responses with the other three parameters. However, no studies have clarified whether and how the CAN algorithm can solve the limitations of the CNI model. The present study systematically uncovers the limitations of the CNI model and the solutions provided by the CAN algorithm: (a) the CNI model does not consider negative values of the parameters, but the CAN algorithm does; (b) the sequential processing assumption of the CNI model is biased, the CAN algorithm proposes a parallel calculation strategy to fix this problem; (c) the calculation of the I parameter of the CNI model is inaccurate, so the CAN algorithm proposes the A parameter to replace it; (d) the CNI model has a problem measuring perverse responses, while the CAN algorithm develops three parameters to measure these. We examined some of our points on the basis of a reanalysis of the foreign language effect (FLE) by comparing the parameters from the CAN algorithm with those from the CNI model. We found that consequence and norm sensitivity were estimated to be greater using the CNI model than with the CAN algorithm. Consequently, these overestimations significantly (consequence sensitivity) and marginally (norm sensitivity) interfered with the FLE, making the FLE more likely to return a false positive result. In addition, the CAN algorithm was able to measure the extent of perverse responses, indicating that foreign language (compared to a native language) leads to more perverse responses. The present study demonstrates that the CNI model magnifies the Type I error of conclusions and that the CAN algorithm (compared to the CNI model) provides more insights regarding moral decision making.