Developing personalized persuasive technologies using an adaptive and data-driven approach

说服性技术(PT)是一种交互系统，它通过手机游戏、手机应用和社交媒体等平台激励和帮助人们采取新的行为。现有的PT是使用个性化模型开发的，这些模型在与系统接触时不适应用户的当前情况。我最近的研究表明，使用个性化说服模式开发的PT比非个性化说服模式具有更好的效果。理解技术如何影响行为是人机交互的首要任务。通过机器学习算法准确反映个人的人类行为仍然是一项悬而未决的挑战。我的长期愿景是开发数据驱动的可信PT，能够增强用户的能力，并激励他们实现对他们和社会重要的目标。有效的PT的基础是复杂的算法，目前还没有方法来结合个性化的用户个性化，以适应用户偏好的变化。Discovery Grant研究计划的短期目标包括：1)为PT开发新的自适应数据驱动的个性化模型，2)在电子学习和严肃游戏领域使用新的个性化模型设计、开发和评估个性化说服系统，以及3)在说服系统的设计和开发中建立信任和伦理模型。对于第一个目标，我将使用贝叶斯机器学习算法开发新的个性化模型，以计算当向用户呈现有说服力的消息或可视化时的成功概率。对于第二个目标，我将把这种新颖的个性化模型应用到一个自适应的电子学习平台的设计和开发中，以克服当前电子学习PT的非适应性挫折。我还将使用新颖的个性化模式来开发健康营养的严肃游戏。为了实现第三个目标，我将制定一个道德框架，以增加人们对PT的信任。目前还没有设计合乎道德和值得信赖的PT以加强其使用的PT设计框架。为了创建这个框架，我将开发多个原型系统来模拟各种伦理场景，以确定用户的可信度。通过开发能够以动态个性化方式表征复杂的人机交互的最先进的算法，我的研究将成为全球首批以动态结合用户现状的方式推进个性化PT的公司之一，这将在许多领域具有深远的应用，如电子学习和健康营养。我的研究将产生第一个伦理PT设计框架，可以指导这一领域的未来发展。HQP将获得从研究设计到评估和实施的经验，并将提高他们的数据挖掘和软件开发技能，这对他们未来在学术界或行业的职业生涯是有益的。

Persuasive technologies (PT) are interactive systems that motivate and help people adopt new behaviours through platforms such as mobile games, mobile applications, and social media. Existing PT are developed using personalization models that are not adaptive to the user's current situation while engaging with the system. My recent research indicates that PT developed with personalized models of persuasion have a greater efficacy than non-personalized ones. Understanding how technology can influence behaviour is at the forefront of human-computer interaction. Accurately reflecting individual human behaviour through machine learning algorithms remains an open challenge. My long-term vision is to develop data-driven trusted PT that can empower users and motivate them to achieve goals that are important to them and society. Underpinning effective PT are complex algorithms, and currently, there are no methodologies to incorporate individualized user personalization that are adaptive to changes in user preferences. The short-term objectives of this Discovery Grant research program include: 1) to develop novel adaptive data-driven personalization models for PT, 2) to design, develop and evaluate personalized persuasive systems using the novel personalization models in the domains of e-learning and serious games, and 3) to model trust and ethics in the design and development of persuasive systems. For the first objective, I will develop novel personalization models using Bayesian machine learning algorithms to compute the probability of success of a persuasive message or visualization when presented to a user. For the second objective, I will apply this novel personalization model in the design and development of an adaptive e-learning platform to overcome the non-adaptive setback of current e-learning PT. I will also use the novel personalization model to develop serious games for healthy nutrition. To achieve the third objective, I will develop an ethical framework to increase people's trust in PT. There is currently no PT design framework for designing ethical and trustworthy PT to enhance their use. To create the framework, I will develop multiple prototype systems to simulate various ethical scenarios to identify users' trustworthiness. Through developing state-of-the-art algorithms that can characterize complex human-computer interactions in a dynamic personalized manner, my research will be among the first globally to advance personalized PT in a manner that dynamically incorporates a user's current situation, which has far-reaching applications across numerous domains such as e-learning and healthy nutrition. My research will result in the first framework for ethical PT design which can guide future development in this area. HQP will gain experience from research design to evaluation and implementation and will enhance their data mining and software development skills which are beneficial to their future careers in academia or industry.