Hongyu Zhou
How to reduce the sense of helplessness of residents in the process of using machine decision
Summary
Helplessness is one of the possible negtitive emotions while using machine decision. What is worse, regardless of whether citizens are willing to live in digital environment, government has begun to apply machine decision technique in public management. Understanding the reason of this helplessness is vital for HCI and growing area of machine decision, which works to improve or creates new modes of citizen implement government policy. In this research, I analytically study how to reduce such helplessness brought by machine decision. To take an example, I conduct experiments on the COVID-19 tracing technology to deomonstrate how my findings can help to increase the user uptake. Tracing technology is used in a mobile phone application that manages personal health information through roaming tracking and contact tracking in the pandemic. The generation of the tracing app is based on the real information of users, including their address, health status and travel history. Although the personal information provided to the government is used for better prevention, such control by the tracing app has brought new inconveniences to citizens. The reason for increasing feeling of helplessness is that people do not know what kind of leadership they are being led by, and the system does not use any form of oversight or accountability, which makes it difficult to detect problematic result generated by the system.
Additional info
A key challenge for designers and engineers building tracking applications is to align them with the values of the public health surveilled, while also communicating these values and the reasons behind the monitoring technique to promote autonomous endorsement. Similarly, psychology can provide evidence-based guidance for the design of autonomy-supported behavioural interventions. Another helpful starting point is frameworks for responsible innovation or ethical AI that are broadly aligned with the principles of biomedical ethics, particularly those supporting human autonomy.
Allen and Ferrand (1999) looked at the impact on three of Geller’s factors predicting actively caring (self-esteem, personal control, and belonging). Only one of these predictor variables, personal control, showed a significant relationship.
So which factor would affect the design of the tracing technology or AI tools needed to control future pandemics by reducing the sense of helplessness for people who have little computer knowledge?
Method
I prepare a background introduction describing the topic and 10 different questions for each group asking their travel and physical health during the pandemic according to their age and gender. Subjects will be randomly allocated to see one of four versions of the app, displayed below. I identified the need and finalized the questions into one pre-test and three post-tests to reduce the participant's reading time to improve their patience, and to make sure that the questions conveyed our meaning concisely to the reader in the shortest possible space.
Environment
Since the beginning of this year, the Corona Virus Disease 2019 (COVID-19) has spread around the world, with more than 200 countries and regions having confirmed cases. As of May 5, the cumulative number of confirmed cases exceeded 3.7 million and the number of deaths exceeded 250,000. Due to the impact of the epidemic, data from the International Monetary Fund (IMF) indicated that the impact was confirmed as the worst economic recession since the Great Depression in the 1930s. At the same time, the World Health Organization(WHO) claims that the death rate of COVID-19 is about 10 times that of influenza.
In this study, I propose to use quantitative research for adults who have use health code and explore three main factors (which could influence citizens’ confidence in AI service for a pandemic). A certain degree of effort space will affect citizens’ perception of whether they have autonomy, which will affect citizens’ positive understanding of the process of using health codes, and affect citizens’ understanding of the process of producing results.
I hope to use 2 x 2 covariance factor analysis (ANCOVA) to find the primary effect of message framing, so that participants reported more positively about the application than the control group under targeted behavioral planning. In addition, by measuring the knowledge of interaction patterns, I can test the independent influence of the experiment on the expected variables, thus controlling the positive cognition and intention of using the health code. Again, I want to find the main role of information processing.
Study 1
Pre-Test Items
Provided participants with a health code page that included the experimental variable version for trial use. What I expect participants to experience when using health code will be an important predictor of test participants’ understanding of machine decision making policies. Therefore, to control for the initial intent, I posed the question "do you think the color identity provided by the health code is acceptable to you? ". The item was answered on a scale of 0(highly unlikely) to 10(highly probable).
Fig. 1. The procedure of how I conduct survey in study 1
Post-Test Items
I posed three post-test questions to assess participants’ views on three levels of color decision status: 1) do you think the color status provided by the health code is acceptable to you? (0 = absolutely impossible; 10 = extremely likely); 2) are you willing to accept a future machine audit similar to the health code to determine your daily life? (0 = Never; 10 = completely predictable) ; 3) would you recommend a health code-like approach involved machine decision-making to other; would you suggest to apply a health code-like approach in machine decision-making; would you persuade your friends to use the health code-like machine decision-making approach friends? (0 = completely impossible;10 = very acceptable). I also asked participants five other questions related to their assessment of the application, trust in the application, perceived technical availability, and their own or other reasons for using the application.
Table 1. Inter-correlations, means, and standard deviations for the variables in Study 1 Note. *p < .05, **p < .01, ***p < .001. Pre-test = pre-experiment; Post-test = post-experiment.
Table 2. Experimental group-specific means and standard deviations for the pre-experiment (pre-test) and post-experiment (post-test) measures in Study 1.
Discussion
The aim of Study1 was to assess two elements of framework of message presentation, and their effects on people’s intentions to use health code and trust in decision-making process. I found support for our hypotheses regarding reduce feelings of helplessness, indicating that the color changes are clearly expressed and citizens are given a certain control space may be vital tools in promoting the uptake of Health code. However, I find an good effect of clear message presentation and the same effect of a certain control space was not as positive as prediction of health code uptake. Participants in the two farming conditions did not differ any outcome. Meanwhile, individuals in the high planning behavior support conditions reported a greater likelihood of using health code and of recommending it to friends and family, compared to those in the low planning behavior support condition. However, information safety did not affect people’s perceptions of the COVID-19 tracing application as a worthwhile use of government resources. The effects of information safety assurances were evident regardless of message framing condition.
Study1 leaves some possibilities not promoted, and thus requires expansion. I using the result of study1 as an precondition to find a way to enhancing citizen’s autonomy and using intention. Thus, in addition to attempting to understanding public use intention, I also tested the links between message framing and use intention of reward interaction mechanism.
