Editing Persuasive technology (section)

== Research challenges ==
Despite the promising results of existing persuasive technologies, there are three main challenges that remain present.

=== Technical challenges ===

Persuasive technologies developed relies on self-report or automated systems that monitor human behavior using sensors and pattern recognition algorithms. Several studies in the medical field have noted that self-report is subject to bias, recall errors and low adherence rates. The physical world and human behavior are both highly complex and ambiguous. Utilizing sensors and machine learning algorithms to monitor and predict human behavior remains a challenging problem, especially that most of the persuasive technologies require just-in-time intervention.

=== Difficulty in studying behavior change ===

In general, understanding behavioral changes require long-term studies as multiple internal and external factors can influence these changes (such as personality type, age, income, willingness to change and more). For that, it becomes difficult to understand and measure the effect of persuasive technologies.
Furthermore, meta-analyses of the effectiveness of persuasive technologies have shown that the behavior change evidence collected so far is at least controversial, since it is rarely obtained by Randomized Controlled Trials (RCTs),{{sfn|Hamari|Koivisto|Pakkanen|2014}} the “gold standard” in causal inference analysis. In particular, due to relevant practical challenges to perform strict RCTs,{{sfn|Bhushan|Steg|Albers|2018}} most of the above-mentioned empirical trials on lifestyles rely on voluntary, self-selected participants. If such participants were systematically adopting the desired behaviors already before entering the trial, then self-selection biases would occur. Presence of such biases would weaken the behavior change effects found in the trials. Analyses aimed at identifying the presence and extent of self-selection biases in persuasive technology trials are not widespread yet. A study by Cellina et al. on an app-based behavior change trial in the mobility field found evidence of no self-selection biases.{{sfn|Cellina|Vittucci Marzetti|Gui|2021}} However, further evidence needs to be collected in different contexts and under different persuasive technologies in order to generalize (or confute) their findings.{{citation needed|date=August 2021}}

=== Ethical challenges ===
The question of manipulating feelings and desires through persuasive technology remains an open ethical debate. User-centered design guidelines should be developed encouraging ethically and morally responsible designs, and provide a reasonable balance between the pros and cons of persuasive technologies.{{sfn|Ijsselsteijn|De Kort|Midden|Eggen|2006}}

In addition to encouraging ethically and morally responsible designs, Fogg believes education, such as through the journal articles he writes, is a panacea for concerns about the ethical challenges of persuasive computers.{{sfn|Fogg|1998}} Fogg notes two fundamental distinctions regarding the importance of education in engaging with ethics and technology: "First, increased knowledge about persuasive computers allows people more opportunity to adopt such technologies to enhance their own lives, if they choose. Second, knowledge about persuasive computers helps people recognize when technologies are using tactics to persuade them."{{sfn|Fogg|1998}}

Another ethical challenge for persuasive technology designers is the risk of triggering persuasive backfires, where the technology triggers the bad behavior that it was designed to reduce.{{sfn|Stibe|Cugelman|2016}}