Tim Maher Gives Lecture on Ambient Intelligence

On Thursday 1 August, GCRI hosted an online lecture by Tim Maher entitled ‘Ambient Intelligence: Implications for Global Environmental Change and Totalitarianism Risk’ (see the pre-lecture announcement). Maher is a recent graduate of Bard College’s M.S. program in Climate Science and Policy and a GCRI Research Assistant. The lecture is based on Maher’s M.S. thesis ‘Ambient Intelligence and Ambient Persuasive Technology: Sustainability and Threats to Autonomy’ [1]. The lecture included discussants Maurits Kaptein, Assistant Professor of Statistics and Research Methods at the University of Tilburg and founder of PersuasionAPI, and Arden Rowell, GCRI Research Associate and Associate Professor at the University of Illinois College of Law.

Maher’s lecture began with an overview of the two types of technological advancements under discussion: Ambient Intelligence (AmI) and Ambient Persuasive Technology (AmPT). AmI is the embedding of smart technology into the built environment. This technology can sense information, process information, and communicate this information with each other and with the Internet. It models individual preferences and behaviors, and maximizes an individual’s comfort by manipulating the surrounding built environment. The AmI in a home would sense, for example, preferences in temperature or lighting and manage those things automatically for the individual. Many examples of such embedding already exist or are in development [2]. AmI can help the environment by doing things like turning off appliances when we don’t need them or adjusting the thermostat for us. Additionally, AmI can effectively automate demand-side management of the electricity grid. This could provide reductions in peak load, the need for new power plants, and concurrent reductions in greenhouse gas emissions.

AmPT uses the same types of technologies to automate the persuasion of behavior. AmPT models individual decision-making processes and then manipulates the surrounding environment to increase the likelihood of individuals taking a specific choice.. One example discussed was “The Shower Calendar” [3], in which a display is projected onto the shower wall showing personal water consumption. A personalized colored dot presented in a calendar-like matrix gets smaller with consumption, thus communicating an individual’s water use compared to others’ in their household, and the limitedness of the resource. The intention behind such a technology is to provide individuals with feedback about their relative water consumption and persuade them to reduce water usage. Because they can be so ubiquitous, AmI and AmPT can reduce our environmental impact through the many big and little environmentally impacting things that we do.

The AmI and AmPT paradigms, Maher explains, are borne out of the advancement of types of enabling technologies such as sensors, microprocessors, and WiFi chips, which are getting smaller, and thus increasingly capable of being embedded directly into everyday material. The repercussions of the extensive integration of these sorts of technologies is difficult to predict. There can be large benefits to these technologies, such as helping society achieve a more sustainable use of resources, increasing economic energy efficiency (through AmI) or increasing the likelihood of behaviors that better the environment (AmPT). However, there are also huge risks, such as issues of individual and state security, privacy, totalitarianism, paternalism, and other threats to autonomy.

The current trajectory of technology suggests that the integration of AmI and AmPT into society is highly likely. How then, Maher asks, do we create policies that enable the benefits while also reducing the risks? At the heart of this is the issue of paternalism, in which the technology designers decide what is best for the users: what ambience to seek, which persuasions to make. But all technologies incorporate their designers’ values. We can’t avoid it. If technological paternalism is inevitable, then perhaps the question is not whether or not they should be developed, but how they are developed, by who, and with what set of values?

Maher proposes a solution. He suggests that in order to increase the benefits of these technologies, individuals should be able to determine what functions or operations AmI and AmPT systems serve within their life, under what circumstances they agree to allow automated ambient systems into their lives, and what information is collected and acted upon. In other words, individuals should be their own paternalists. What does this mean? This means, in theory, that individuals should be able to decide under what circumstances they agree to be ambiently persuaded, if ever, and what information is persuasively transmitted.

Online participants raised several points about this idea. To some, there remain major grey areas in the boundaries between nudging an individual toward a certain behavior (say, by highlighting a particular option with a different color in a text, or making one option the default that an individual would need to opt-out of) and actually coercing an individual. One discussant suggested the role of the Fundamental Attribution Error – a concept in social psychology in which people tend to think that they themselves are not impacted or influenced by ‘nudges’, but believe other people are. If this holds true for different types of AmI in the context of evaluating people’s ability to engage with these technologies, one concern is that many people will not manage their own AmI settings, and those that do will not fully understand how the settings are impacting their behavior.

Participants were also interested in the government’s role in private policy design and implementation. One discussant pointed out that there may be existing agencies whose purview could, or should, include this sort of monitoring. The Federal Communications Commission is one agency whose purview already includes expanding and strengthening the nation’s communication infrastructure, and which could arguably be well suited for monitoring and regulating this sort of technology. Recent policy developments relate to these issues: under Executive Order 13563, “U.S. regulators, to the extent permitted by law, to select approaches that maximize net benefits; choose the least burdensome alternative; increase public participation in the rulemaking process; design rules that are simpler and more flexible, and that provide freedom of choice; and base regulations on sound science” [4]. The idea of increased public participation in the rulemaking process and design rules that are flexible is being mandated at the executive level, which could have implications for AmI and AmPT policy.

Overall, the issue of informed consent was central to this discussion. How informed consent is accomplished, and in what way people’s autonomy is translated into transparent, flexible policies remains a complex and compelling issue for the times ahead.

A full abstract of the online lecture is available here:

One set of emerging information and communication technologies, Ambient Intelligence (AmI) and Ambient Persuasive Technology (AmPT), offers large possible benefits toward encouraging a more sustainable use of resources. Ambient Intelligence can increase economic and energy efficiency by automating the built environment. Ambient Persuasive Technology can help increase the likelihood of behaviors that better the environment. However, these technologies come with great risk. In this presentation, I analyze this risk, focusing on issues of individual and state security, privacy, and totalitarianism, and conclude that the majority of this risk can be distilled down to the threat of paternalism and other threats to autonomy. I then analyze the ethics of AmI and AmPT paternalism and potential threats to autonomy using Kantian deontological and Millian consequentialist frameworks. Assuming this set of ethical views is correct, this ethical investigation concludes that the only moral way to implement these technologies is to give individual users the ability to control their relationship with AmI and AmPT. In effect, users should be their own paternalists. I conclude with discussions and policy recommendations for improving data security, enabling users to be more involved in the design process of AmI and AmPT, and enabling users greater direct control over their own levels of consent for each AmI and AmPT operation.

The presentation was hosted online via Skype, with slides shown on PowerPoint. The attendees included Gautam Sethi, Associate Professor of Economics at the Bard College Center for Environmental Policy, Miles Brundage, PhD student at Arizona State University’s Human and Social Dimensions of Science and Technology, and GCRI’s Seth Baum, Tony BarrettKaitlin Butler, Grant Wilson, Mark Fusco, and Robert de Neufville.

Thanks to Maurits Kaptein, Assistant Professor of Statistics and Research Methods at the University of Tilburg and founder of PersuasionAPI, and Arden Rowell, GCRI Research Associate and Associate Professor at the University of Illinois College of Law for their role as discussants in this lecture.

[1] Maher, Tim (May 2013). Ambient Intelligence and Ambient Persuasive Technology: Sustainability and Threats to Autonomy. Bard Center for Environmental Policy: Annandale on Hudson, NY.

[2] Wasik, Bill (2013). Welcome to the Programmable World. Wired Magazine.

[3] Laschke, Matthias, et al. (2011). “With a little help from a friend: a shower calendar to save water.” CHI’11 Extended Abstracts on Human Factors in Computing Systems. ACM.

[4] Sunstein, Cass (2012). Reducing Red Tape: Regulatory Reform Goes International. The Office of Management and Budget, White House.

This post was written by

2 Comments on "Tim Maher Gives Lecture on Ambient Intelligence"

  • Usama Anber says

    Thanks for posting on such an important topic. I have some comments and questions below, and please excuse my poor psychology background.

    My understanding is that AmI and AmPT are technologies used to optimize energy consumption in such a way to maximize energy efficiency either directly, AMI, or indirectly through individuals’ behavior, AmPT. Hence, I am not sure how the concept of paternalism (which is a form of control if I understand it correctly) fits into the discussion? Why is it needed (even if the users should be their own paternalists)? Examples?

    Also, while incorporating AmI and AmPT technologies into buildings, businesses, transportation, … etc. is very conceivable; I don’t see how it might be perilous to state security or how sharing such data could be of potential threat by terrorists or hackers. Would you please explain? Examples?

    Finally, is there a model study of how using AmI and AmPT might impact the collective behavior of a society (rather than individual), and the energy consumption-efficiency in a city like New York in the next 10-20 years?


  • Tim Maher says


    Thank you for your interest and comments. I will do my best to answer your questions in a succinct manner.

    It is true that one possible application of these technologies is to automate the built environment and maximize energy efficiency (AmI) or to increase the likelihood of behaviors that better the environment (AmPT). However, there are many other possible applications of these technologies, especially with AmPT. Once AmPT understands how you make decisions, it can essentially attempt to make you more likely to take any specific choice it deems worthy. In short, what these technologies do is constantly surveil individuals, collecting biometric and behavioral data in order to estimate/predict their personal preferences, behavior, and even their decision-making processes. These technologies then use this individualized knowledge to automatically make decisions and take actions that alter the user’s surrounding environment. The end result is that the actions taken by AmI and AmPT may be paternalistic actions. In order for an action to be considered paternalistic, it must meet four conditions: the interference condition, the consent condition, the benevolence condition, and, sometimes, the superiority condition. AmI and AmPT actions can automatically make judgments and take actions (interference) on behalf of an individual (benevolence) without their consent (consent), implying that these technologies know better than the individual of what is best for the individual (superiority). All of this can be accomplished without the individual’s knowledge.

    Those charged with designing, regulating, and maintaining AmI and AmPT systems wield great power and influence over society, and it is important that this power not be abused by public, private, or independent parties. Corporate abuse could result in overly invasive marketing. Government abuse could result in unwelcome monitoring and oppression of its citizens or new forms of geopolitical cyber warfare between states. And AmI and AmPT systems are also vulnerable to malicious independent rogue agents interested in stealing information or perhaps even hijacking AmPT systems in order to influence individuals towards some attitude or behavior change that could benefit the rogue agent.

    These risks – individual and state security, privacy, totalitarianism – are all, in effect, threats to individual liberty and autonomy. If data security is compromised, individuals would lose control over their own information and may even be influenced without their consent. Unbridled AmI and AmPT systems threaten individual rights to privacy and may even result in a surveillance society that oversteps the boundary between public and private life. Additionally, AmI and AmPT introduce a new frontier for cyber warfare, whereby individuals could become pawns used against their own nation (covert operations focused on accessing AmPT profiles and influencing the members of rival states – the evolution of spying and meddling). AmI and AmPT could also provide the tools necessary for totalitarian regimes to more easily control their population. This could result in a higher likelihood of states transitioning to totalitarianism in addition to totalitarian states becoming more stable. Distilling these threats down to threats of paternalism and autonomy allows for an ethical investigation of these technologies. Navigating the ethics of these technologies is essential to the design of appropriate policies to address AmI and AmPT.

    Regarding your last question, I would like to note that the AmI and AmPT paradigms are still mostly in their infancy. It will likely be a few decades until these technologies are completely ubiquitous. This is good because it gives us more time to develop sufficient policies before the worst-case scenarios can come to fruition! That being said, there are several pilot and/or research projects focused on the intersection of energy efficiency and AmI. Here is a short list to get you started:

    Alam, M., Reaz, M., and Ali, M. (2011). A Spatiotemporal Model Of Human Circadian Rhythym In Smart Homes. Applied Artificial Intelligence, 25(9), 788–798.
    Ham, J., and Midden, C. (2010). Ambient Persuasive Technology Needs Little Cognitive Effort: The Differential Effects of Cognitive Load on Lighting Feedback versus Factual Feedback. In Ploug, T., Hasle, P., and Oinas-Kukkonen, H. (Eds.), Persuasive Technology, 6137, 132–142.
    Ham, J., Midden, C., and Beute, F. (2009). Can ambient persuasive technology persuade unconsciously?: using subliminal feedback to influence energy consumption ratings of household appliances. In Proceedings of the 4th International Conference on Persuasive Technology, 29.
    Higginson, S., Richardson, I., and Thomson, M. (2011). Energy use in the context of behaviour and practice: The interdisciplinary challenge in modelling flexible electricity demand. Presented in the proceedings of Energy and People: Futures, Complexity and Challenges, Oxford University.
    Moura, P. S., and de Almeida, A. (2010). The role of demand-side management in the grid integration of wind power. Applied Energy, 87(8), 2581–2588. doi:10.1016/j.apenergy.2010.03.019
    Mozer, M. (2005). Lessons from an adaptive home. Smart Environments, 271–294.
    Mozer, M. (2008). 12 Lessons from an Adaptive Home.
    Ramchurn, S., Vytelingum, P., Rogers, A., and Jennings, N. (2011). Agent-based control for decentralised demand side management in the smart grid. In Proceedings of The 10th International Conference on Autonomous Agents and Multiagent Systems, 1, 5-12
    Strbac, G. (2008). Demand side management: Benefits and challenges. Energy Policy, 36(12), 4419–4426.

    I hope you find this helpful.

    My best,