AUgust 2019
Designing the Invisible
An Introduction to Anticipatory Design
Anticipatory design is the premise of reducing users' cognitive overload by facilitating their decision-making process on behalf of them, especially in the Pre-interaction timeframe.
Abstract
In 2004, a highly influential book, The Paradox of Choice, was published by psychologist Barry Schwartz. This book was a compelling manifesto that outlined the effects of abundance and choice in a person's life. A decade later, Aaron Shapiro, the CEO of the global digital design agency Huge, developed Schwartz's findings into a new scenario, that he coined as Anticipatory Design on a publication for the Fast Company magazine . Schwartz defend that too many choices lead to poor quality decisions and less satisfied users. Now, the recommended solution for this problem might relapse on this recent approach called Anticipatory Design.
Keywords
Anticipatory Design, Artificial Intelligence, Interaction Design, Decision-making
Introduction
There is no novelty in the term. Anticipatory comes from the Latin verb Anticipalis which means to "anticipate, get the lead, get ahead of; have preconception; occupy beforehand." . The term "anticipatory" appears in previous studies in computer science (Zamenopoulos, 2007), philosophy (Husserl, 1991), information science (Zamenopoulos, 2007), interaction design (Van Bodegraven, 2017), and even in biology (Rosen, 1985). All the fields discuss anticipation as a notion of predicting future actions as essential and inherent components of systems design.
Anticipatory design principles have their root on anticipation and, according to Poli (2009), the best definition of anticipation came from mathematician Rosen's,
Anticipatory design principles have their root on anticipation and, according to Poli (2009), the best definition of anticipation came from mathematician Rosen's,
An anticipatory system is a system containing a predictive model of itself and/or its environment, which allows it to change at an instant in accord with the model's predictions pertaining to a later instant. (Poli, 2009 p. 2)
Anticipatory Design is the premise of reducing users' cognitive overload by facilitating their decision-making process on behalf of them, especially in the Pre-interaction timeframe.
Although, Anticipatory design is not triggered in pre-events, it also evoked by post-events. That is why designing for anticipatory design can get the experience stuck in an Experience Bubble:
Although, Anticipatory design is not triggered in pre-events, it also evoked by post-events. That is why designing for anticipatory design can get the experience stuck in an Experience Bubble:
Your filter bubble is your own personal, unique universe of information that you live in online. And what's in your filter bubble depends on who you are, and it depends on what you do. But the thing is that you don't decide what gets in. And more importantly, you don't actually see what gets edited out. (Pariser, 2011)
Pariser argues that algorithms create filter bubbles that are affecting our serendipity.
Notwithstanding, the future designer role under anticipatory design will be to envision environments that eliminate as many steps of interaction as possible to simplify the service processes. This will mean that users will not have to route throughout several options in each stage of a service. Instead, smart algorithms, if possible, will be able to make most of the decisions for them.
To the designer, the critical factor here will be to acknowledge which level (by criticality) of decisions the system will automate. The designer should design criticality decisions levels that the system should take into consideration on behalf of the user. They will need to know when to make decisions for their users and when to let them be the deciders (Doody, 2018).
Van Bodegraven (2017, p. 438) differentiates anticipatory design as a pattern and not as a method. To him, the method is Predictive UX. Here, the designer can focus on the service design and support the service with anticipatory design as a design pattern. Accordingly, personalization helps users make a decision; in contrast, anticipation chooses on behalf of the user. The anticipatory design falls into two solutions models: One Closed-word AI, where the algorithms are designed to help users to decide, and second, Open-world AI, where the algorithms are designed to help users to discover (e.g., Spotify or Netflix) (Chen, 2016).
Notwithstanding, the future designer role under anticipatory design will be to envision environments that eliminate as many steps of interaction as possible to simplify the service processes. This will mean that users will not have to route throughout several options in each stage of a service. Instead, smart algorithms, if possible, will be able to make most of the decisions for them.
To the designer, the critical factor here will be to acknowledge which level (by criticality) of decisions the system will automate. The designer should design criticality decisions levels that the system should take into consideration on behalf of the user. They will need to know when to make decisions for their users and when to let them be the deciders (Doody, 2018).
Van Bodegraven (2017, p. 438) differentiates anticipatory design as a pattern and not as a method. To him, the method is Predictive UX. Here, the designer can focus on the service design and support the service with anticipatory design as a design pattern. Accordingly, personalization helps users make a decision; in contrast, anticipation chooses on behalf of the user. The anticipatory design falls into two solutions models: One Closed-word AI, where the algorithms are designed to help users to decide, and second, Open-world AI, where the algorithms are designed to help users to discover (e.g., Spotify or Netflix) (Chen, 2016).
Anticipatory Design is the bridge that links algorithm-powered AI with user-centric design disciplines behind this technological world.
In sum, the design field is moving forward from empowering users with tools to help them make many decisions at each step of the process to serving users by providing them smart services that are continually learning where and how to make peoples' lives easier by sliding AI technology into the background of their perceptions.
Another author, Doody (2018), makes an interesting distinction: to her, anticipatory design focuses on predicting people's needs and helping them make better choices. Whether making decisions on behalf of the user, she calls it Automate Design instead of anticipatory design. The original definition of anticipatory design was a mix between the two, but Doody makes a relevant distinction that can solve some of the current risks brought by anticipatory design that we will cover up further on this paper.
Another author, Doody (2018), makes an interesting distinction: to her, anticipatory design focuses on predicting people's needs and helping them make better choices. Whether making decisions on behalf of the user, she calls it Automate Design instead of anticipatory design. The original definition of anticipatory design was a mix between the two, but Doody makes a relevant distinction that can solve some of the current risks brought by anticipatory design that we will cover up further on this paper.
Anticipatory Design
Deconstructing the Principles
This design method moves around three major concepts:
This design method moves around three major concepts:
Figure 1: Anticipatory Design structure retrieved from Van Bodegraven (2017)
If one of these three actors fail, we cannot design for anticipatory design. They need to be aligned and effectively used (Van Bodegraven, 2017) . "The goal is not to help the user make a decision but to create an ecosystem where a decision is never made" (Shapiro, 2015).
With this symbiosis among IoT, ML and UXD, Van Bodegraven (2017, p. 14) concludes that "smart technology learns within the IoT by observing, while data is interpreted by machine learning algorithms. Along the way, UXD is crucial for delivering a seamlessly anticipated experience that takes users away from technology."
Those are interesting points in view, especially if we envision these developments, for instance, in the e-commerce industry. For example, how will anticipatory design change or shape people's buying patterns and how will it change their purchase routines or behaviors by making and eliminating users' choice along the way. According to Shapiro (2015), the goal is always to eliminate as many steps as possible. To do so, AI will need to support anticipatory design with finding ways to use data, prior behaviors and business logic to have a structured ecosystem to automatically decide successfully on behalf of the user or as close as possible.
For Clark, "a systems model built with anticipatory design principles is speculating about user needs and attempting to fill in the blanks correctly" (2016). To him, given these ideas of speculation and prediction, the anticipatory design is not without its critics. In an excellent case in point, we share the same apprehension for anticipatory design. Zamenopoulos and Alexiou (2007) as well as Clark (2016) they made the reservation that when one enters into modes of prediction, we introduce constraints to understanding as well as serendipity since the system could be unable us the capacity of making fortunate discoveries by chance (Van Allen, 2017). For Anne Quito (2015) anticipatory design "is a radical shift in terms of thinking about design." These remarks equally apply to the importance of developing a heuristic system focus on the Pre-interaction experiences that have in mind fairness, accountability, interpretability, and transparency of the service (Riedl, 2019).
Services like Google Now are one of the successful examples of what anticipatory systems are starting to be capable of. This application follows the anticipatory design principles of making their life easier and better by reducing and simplification the number of users' inputs and decisions. Google Now is trained to predict when a user is about to take certain actions and offers help accordingly.
With this symbiosis among IoT, ML and UXD, Van Bodegraven (2017, p. 14) concludes that "smart technology learns within the IoT by observing, while data is interpreted by machine learning algorithms. Along the way, UXD is crucial for delivering a seamlessly anticipated experience that takes users away from technology."
Those are interesting points in view, especially if we envision these developments, for instance, in the e-commerce industry. For example, how will anticipatory design change or shape people's buying patterns and how will it change their purchase routines or behaviors by making and eliminating users' choice along the way. According to Shapiro (2015), the goal is always to eliminate as many steps as possible. To do so, AI will need to support anticipatory design with finding ways to use data, prior behaviors and business logic to have a structured ecosystem to automatically decide successfully on behalf of the user or as close as possible.
For Clark, "a systems model built with anticipatory design principles is speculating about user needs and attempting to fill in the blanks correctly" (2016). To him, given these ideas of speculation and prediction, the anticipatory design is not without its critics. In an excellent case in point, we share the same apprehension for anticipatory design. Zamenopoulos and Alexiou (2007) as well as Clark (2016) they made the reservation that when one enters into modes of prediction, we introduce constraints to understanding as well as serendipity since the system could be unable us the capacity of making fortunate discoveries by chance (Van Allen, 2017). For Anne Quito (2015) anticipatory design "is a radical shift in terms of thinking about design." These remarks equally apply to the importance of developing a heuristic system focus on the Pre-interaction experiences that have in mind fairness, accountability, interpretability, and transparency of the service (Riedl, 2019).
Services like Google Now are one of the successful examples of what anticipatory systems are starting to be capable of. This application follows the anticipatory design principles of making their life easier and better by reducing and simplification the number of users' inputs and decisions. Google Now is trained to predict when a user is about to take certain actions and offers help accordingly.
Anticipatory Design Context
Designers and data scientists both have a responsibility to understand how to shape experiences that improve lives (Girardin and Lathia, 2017). However, the design value comes from Quality of Experience [1]. Designers cannot improve people's quality of life if they do not achieve any utilitarian and hedonic value.
How designers will define if a particular product or service is ready for anticipatory design will need to follow specific criteria. One of the fundamental criteria is if a particular product or service has to have an online component that is collecting data from individuals. As we saw previously, there is no anticipatory design if it not gathers together three factors: IoT, ML, and UXD (Van Bodegraven, 2017; Doody, 2018).
Accordingly, the collection process can be built in several ways. Through an actual physical product (IoT like Nest, Alexa, wearables, etc.) that can be sitting collecting data from individuals locally or can be integrated into people's phones. Another criterion lies in preparation for collecting massive amounts of data or buy access to it. A large business can generate vast amounts of data, but for those smaller businesses or young services or products generate large amounts of data can become a problem. In that sense, they can resort to purchasing generalized data, and they create personas [2] that are similar to the smaller user sample. The mandatory for process to anticipatory is to have data to work upon, the strategies to gather that data can be diverse. The ultimate goal is to collect attributes of a person, whether individually or collectively, to generate sustained data to customize experiences. Currently, the anticipatory design brings to the table three advantages. Reducing the cost of choices, simplification of the user interface (UI), and at last but not least, improving the quality of the decision-making process.
How designers will define if a particular product or service is ready for anticipatory design will need to follow specific criteria. One of the fundamental criteria is if a particular product or service has to have an online component that is collecting data from individuals. As we saw previously, there is no anticipatory design if it not gathers together three factors: IoT, ML, and UXD (Van Bodegraven, 2017; Doody, 2018).
Accordingly, the collection process can be built in several ways. Through an actual physical product (IoT like Nest, Alexa, wearables, etc.) that can be sitting collecting data from individuals locally or can be integrated into people's phones. Another criterion lies in preparation for collecting massive amounts of data or buy access to it. A large business can generate vast amounts of data, but for those smaller businesses or young services or products generate large amounts of data can become a problem. In that sense, they can resort to purchasing generalized data, and they create personas [2] that are similar to the smaller user sample. The mandatory for process to anticipatory is to have data to work upon, the strategies to gather that data can be diverse. The ultimate goal is to collect attributes of a person, whether individually or collectively, to generate sustained data to customize experiences. Currently, the anticipatory design brings to the table three advantages. Reducing the cost of choices, simplification of the user interface (UI), and at last but not least, improving the quality of the decision-making process.
[1] Quality of Experience measures the delight or annoyance of a user when experiencing a service. It captures people's aesthetic and hedonic needs.
[2] A persona is user experience design tool to document certain type of users. Personas are fictional, but representative, profiles of target users. It will describe the ways in which certain types of people will use your product or service. Usually one persona is created for each type of user. Personas are used to show us the motivations, pain points and goals that users will be trying to achieve on a certain product or service (Cooper, 1999).
[2] A persona is user experience design tool to document certain type of users. Personas are fictional, but representative, profiles of target users. It will describe the ways in which certain types of people will use your product or service. Usually one persona is created for each type of user. Personas are used to show us the motivations, pain points and goals that users will be trying to achieve on a certain product or service (Cooper, 1999).
Reducing the Cost of Choice
Reducing the cognitive load and decision fatigue that occurs by too much information or too many options, especially when it happens all at once. To Schwartz (2004), providing too many options will make the decision-making process overwhelming and stressful to the users. Because too many choices require a load in the mental effort to uncover and decide something. When this happens in digital services, people tend to leave early, and those who stay are usually less satisfied with the overall user experience.
To avoid this, the best solution is to maximize the experience by minimizing the cognitive load. Schwartz (2004) defended that it is better to have fewer choices than it is to have more. So, the designer should make an effort to eliminate redundant choices by focusing on anticipatory design to achieve a reduction in decision fatigue. There are two essential effects driven by decision fatigue. On the one hand, it reduces the ability to make trade-offs, while it also contributes to decision avoidance (Anderson, 2003).
If an individual is mentally exhausted, one tends to become reluctant to make trade-offs, which might result in perform poor choices. The cause of decision fatigue cannot result in decision avoidance, where the individual does not make any decision at all (Anderson, 2003).
And we must not forget to diagnose that when we are mental fatigue, we tend to resort to mental shortcuts to make decisions quickly and effectively. By doing this, we may follow in a series of harmful cognitive bias that can compromise our decision-making process. Decision fatigue can compromise our ability to prioritize the importance of the trade-off accurately.
Reducing the cognitive load and decision fatigue that occurs by too much information or too many options, especially when it happens all at once. To Schwartz (2004), providing too many options will make the decision-making process overwhelming and stressful to the users. Because too many choices require a load in the mental effort to uncover and decide something. When this happens in digital services, people tend to leave early, and those who stay are usually less satisfied with the overall user experience.
To avoid this, the best solution is to maximize the experience by minimizing the cognitive load. Schwartz (2004) defended that it is better to have fewer choices than it is to have more. So, the designer should make an effort to eliminate redundant choices by focusing on anticipatory design to achieve a reduction in decision fatigue. There are two essential effects driven by decision fatigue. On the one hand, it reduces the ability to make trade-offs, while it also contributes to decision avoidance (Anderson, 2003).
If an individual is mentally exhausted, one tends to become reluctant to make trade-offs, which might result in perform poor choices. The cause of decision fatigue cannot result in decision avoidance, where the individual does not make any decision at all (Anderson, 2003).
And we must not forget to diagnose that when we are mental fatigue, we tend to resort to mental shortcuts to make decisions quickly and effectively. By doing this, we may follow in a series of harmful cognitive bias that can compromise our decision-making process. Decision fatigue can compromise our ability to prioritize the importance of the trade-off accurately.
Simplification of User Interface
Fewer choices on a screen means cleaner and fewer clutter screens. Having fewer choices or no choices at all may seem a natural evolution as voice UI, and AI becomes widespread. As a result, we are moving forward into the simplification of UI. Krishna (2015) argues that in the future, the best interface will have no interface.
The potential for voice UI, motion, and physical interactions [3] are huge, and the trend of physical and digital-connected products will keep blending as fast as IoT keep spreading across industries. In this sense, anticipatory design may have a word to say; it can save much time and user micro-interactions, allowing them to focus on extending the human and creative capabilities, instead of performing repetitive tasks that machines are better than us to perform it (Daugherty and Wilson, 2018).
Fewer choices on a screen means cleaner and fewer clutter screens. Having fewer choices or no choices at all may seem a natural evolution as voice UI, and AI becomes widespread. As a result, we are moving forward into the simplification of UI. Krishna (2015) argues that in the future, the best interface will have no interface.
The potential for voice UI, motion, and physical interactions [3] are huge, and the trend of physical and digital-connected products will keep blending as fast as IoT keep spreading across industries. In this sense, anticipatory design may have a word to say; it can save much time and user micro-interactions, allowing them to focus on extending the human and creative capabilities, instead of performing repetitive tasks that machines are better than us to perform it (Daugherty and Wilson, 2018).
[3] E.g., Project Soli. Soli is a new sensing technology that uses miniature radar to detect touchless gesture interactions. Soli is a purpose-built interaction sensor that uses radar for motion tracking of the human hand. https://atap.google.com/soli/
Improving the Quality of the Decision-making
In this age of ubiquitous computing, algorithms are becoming embedded widely in the world that surrounds us. Smart devices are everywhere, and they are continually gathering our actions, behavior, and preferences. This represents a generation of vast amounts of personal and public data at disposal. With the rising of Big Data and the emergence of expert systems, businesses start to rely more on the efficiency of machines to perform analytics and statistical rather than humans. With this context to support designer's creation process, through anticipatory design, designers can improve the decision-making process and reduce human mistakes by aggregating, gathering, and making use of a lot more data that is manually possible. Designers should predict the user's needs in certain situations to achieve normality and unobtrusiveness of the technology. AI will become an ally in design.
Therefore, we need to improve our understanding of how people accept or reject AI applications to enhance, clarify, and increase their acceptance to it.
Unfortunately, designers are lacking prototyping tools or heuristics for working with AI (Dove et al., 2017). Without proper methods and prototyping tools, it becomes hard to successfully prototype for interactions that may follow unpredictable intelligent evolving courses. Designers will need prototyping frameworks that support anticipatory design and animistic design principles and methods. Otherwise, designers will keep struggling into exploring and exploit the space of possibilities. As such, we need proper tools to overtones the challenges of interacting with unpredictable systems.
The Quality of Experience is not restricted only to technology. And for that reason, we need to improve our understanding of how people accept or reject AI applications, in order to enhance, clarify, and increase their acceptance to it. This will lead us to a significant area of inquiry: How to do anticipatory experience design for AI-driven services?
In this age of ubiquitous computing, algorithms are becoming embedded widely in the world that surrounds us. Smart devices are everywhere, and they are continually gathering our actions, behavior, and preferences. This represents a generation of vast amounts of personal and public data at disposal. With the rising of Big Data and the emergence of expert systems, businesses start to rely more on the efficiency of machines to perform analytics and statistical rather than humans. With this context to support designer's creation process, through anticipatory design, designers can improve the decision-making process and reduce human mistakes by aggregating, gathering, and making use of a lot more data that is manually possible. Designers should predict the user's needs in certain situations to achieve normality and unobtrusiveness of the technology. AI will become an ally in design.
Therefore, we need to improve our understanding of how people accept or reject AI applications to enhance, clarify, and increase their acceptance to it.
Unfortunately, designers are lacking prototyping tools or heuristics for working with AI (Dove et al., 2017). Without proper methods and prototyping tools, it becomes hard to successfully prototype for interactions that may follow unpredictable intelligent evolving courses. Designers will need prototyping frameworks that support anticipatory design and animistic design principles and methods. Otherwise, designers will keep struggling into exploring and exploit the space of possibilities. As such, we need proper tools to overtones the challenges of interacting with unpredictable systems.
The Quality of Experience is not restricted only to technology. And for that reason, we need to improve our understanding of how people accept or reject AI applications, in order to enhance, clarify, and increase their acceptance to it. This will lead us to a significant area of inquiry: How to do anticipatory experience design for AI-driven services?
Anticipatory Design Risks & Challenges
Despite the promising advances brought by this new field, academia has little research regarding the subject. Possible implications under this method have not been appropriately studied. There is a lot to be debated and study under subjects like ethical challenges under design and data, privacy politics, the cost of being right or wrong, the need for new heuristics as well forecast serendipity in automated systems.
Ethical Design
"UX designers are getting more exposed to ethical design since much confidentiality is involved by creating predictive user experiences" (Van Bodegraven, 2017, p. 435). Anticipatory design is resulting in designers getting more involved in areas such as ethical design since much personal data is involved in predictive experiences.
With the rising of fully-autonomous predictive system's designers will face a higher responsibility concerning users' privacy and data collection. Further, if the future will be designing for integrated services, some ethical design issues may arise. Designing for anticipation will obligatorily imply data, and in a ubiquity era, this data and services where the data relies upon, need to be cross-gathered from several other services and resources.
Businesses and even designers will need to understand how to design experiences and services for this cross-functional market. Several industries may need to learn how to work in symbiosis, or openly, for a ubiquitous market service to leverage their customers' experiences and their own business itself. Will businesses be willing to open up to this future sharing cross-functional collaboration? And what ethical concerns will follow into the designer role?
"UX designers are getting more exposed to ethical design since much confidentiality is involved by creating predictive user experiences" (Van Bodegraven, 2017, p. 435). Anticipatory design is resulting in designers getting more involved in areas such as ethical design since much personal data is involved in predictive experiences.
With the rising of fully-autonomous predictive system's designers will face a higher responsibility concerning users' privacy and data collection. Further, if the future will be designing for integrated services, some ethical design issues may arise. Designing for anticipation will obligatorily imply data, and in a ubiquity era, this data and services where the data relies upon, need to be cross-gathered from several other services and resources.
Businesses and even designers will need to understand how to design experiences and services for this cross-functional market. Several industries may need to learn how to work in symbiosis, or openly, for a ubiquitous market service to leverage their customers' experiences and their own business itself. Will businesses be willing to open up to this future sharing cross-functional collaboration? And what ethical concerns will follow into the designer role?
Data Security and Privacy
Algorithms as decision makers will imply new definitions of privacy. Anticipatory design requires context, and context requires data. Anticipatory design is grounded in data, and automation requires a large number of data sets. Data and AI will be the foundation of efficient user experiences for anticipatory systems to work. Designers will need data on users' preferences and previous actions. Without data, intelligent algorithms cannot work, and designers will not be able to design for anticipations.
The issue and concerns are with the new security and privacy policies like the General Data Protection Regulation (GDPR). GDPR is a European new privacy law to protect user's data sharing and gathering. The law does not exist in the United States, but companies who have European customers need to take it into account. If data law restriction continues to grow, it will complicate the design process under anticipatory systems. "A worrying attitude because it may inhibit the development around predictive UX" (Van Bodegraven, 2017, p.436).
Nonetheless, in one hand, designers will need data to predict user behavior, but on the other hand, designers will have to take users privacy matter more into consideration than ever before and incorporate it into the design workflow as transparency mechanism. "Automation will ask much more transparency from its users to estimate needs correctly. The current privacy-ecosystem is not sufficient and scalable in that regard" (Van Bodegraven, 2017, p. 436).
Algorithms as decision makers will imply new definitions of privacy. Anticipatory design requires context, and context requires data. Anticipatory design is grounded in data, and automation requires a large number of data sets. Data and AI will be the foundation of efficient user experiences for anticipatory systems to work. Designers will need data on users' preferences and previous actions. Without data, intelligent algorithms cannot work, and designers will not be able to design for anticipations.
The issue and concerns are with the new security and privacy policies like the General Data Protection Regulation (GDPR). GDPR is a European new privacy law to protect user's data sharing and gathering. The law does not exist in the United States, but companies who have European customers need to take it into account. If data law restriction continues to grow, it will complicate the design process under anticipatory systems. "A worrying attitude because it may inhibit the development around predictive UX" (Van Bodegraven, 2017, p.436).
Nonetheless, in one hand, designers will need data to predict user behavior, but on the other hand, designers will have to take users privacy matter more into consideration than ever before and incorporate it into the design workflow as transparency mechanism. "Automation will ask much more transparency from its users to estimate needs correctly. The current privacy-ecosystem is not sufficient and scalable in that regard" (Van Bodegraven, 2017, p. 436).
Service responsibility
According to Van Bodegraven (2017), if we design a full-autonomous predictive system where all decisions are predictive and anticipated without the user interact with the system or had the opportunity to change a pattern, it may violate the concept of free will. Anticipatory design, as a method, can have repercussions of a dark pattern (Gray, 2018).
In consonance, organizations that are investing in AI and walking into the anticipatory design need to have an introspective attitude and asses responsibility on what they should automate. Just because technology allows to automate something, ethically, it does not mean that they should do it or that the user will want it. There is a line that separates automation versus what decisions users will always want to make. For designers, this line will be the key to designing solutions that understand what level of trust and automation users are willing to delegate (in the decision-making steps) to services fully-autonomous.
It is in this new context that designers need to gain new skills in how to consciously design transparent systems that keep users well-informed and avoid oversimplification so that they feel that they are in control of the technology.
According to Van Bodegraven (2017), if we design a full-autonomous predictive system where all decisions are predictive and anticipated without the user interact with the system or had the opportunity to change a pattern, it may violate the concept of free will. Anticipatory design, as a method, can have repercussions of a dark pattern (Gray, 2018).
In consonance, organizations that are investing in AI and walking into the anticipatory design need to have an introspective attitude and asses responsibility on what they should automate. Just because technology allows to automate something, ethically, it does not mean that they should do it or that the user will want it. There is a line that separates automation versus what decisions users will always want to make. For designers, this line will be the key to designing solutions that understand what level of trust and automation users are willing to delegate (in the decision-making steps) to services fully-autonomous.
It is in this new context that designers need to gain new skills in how to consciously design transparent systems that keep users well-informed and avoid oversimplification so that they feel that they are in control of the technology.
The Cost of Being Wrong
Advances in technologies are thrilling. However, they will fail if they do not address human needs or enhance users' experiences. The anticipatory design comes with certain risks, and that is why we need strategies to mitigate it.
Anticipatory design supported by AI will make an educated guess decision based on user habits and data history. However, what will be the probability of being right and the cost of being wrong? Sophie Kleber, a former member from Huge, the same global digital design agency found by Aaron Shapiro, exposes that the danger of being wrong will be restricted link to the context in which anticipatory design is being applied. The danger of being wrong is relative to each business and its customers. In specific industries, failing with anticipatory design will be a high-risk proposition, as the health of finance for instances. To her, understanding users' needs and expectations is crucial to determining what level of risk are we willing to concern. And on which level users will be able to forgive or trespass such errors.
A lousy song suggested by Spotify is easy to overlook, whereas a wrong decision on health services can lead to significant implications. To help to overcome those conditions, Kleber proposes a diagram to evaluate the impact based on the probability of being right and the cost of being wrong:
Advances in technologies are thrilling. However, they will fail if they do not address human needs or enhance users' experiences. The anticipatory design comes with certain risks, and that is why we need strategies to mitigate it.
Anticipatory design supported by AI will make an educated guess decision based on user habits and data history. However, what will be the probability of being right and the cost of being wrong? Sophie Kleber, a former member from Huge, the same global digital design agency found by Aaron Shapiro, exposes that the danger of being wrong will be restricted link to the context in which anticipatory design is being applied. The danger of being wrong is relative to each business and its customers. In specific industries, failing with anticipatory design will be a high-risk proposition, as the health of finance for instances. To her, understanding users' needs and expectations is crucial to determining what level of risk are we willing to concern. And on which level users will be able to forgive or trespass such errors.
A lousy song suggested by Spotify is easy to overlook, whereas a wrong decision on health services can lead to significant implications. To help to overcome those conditions, Kleber proposes a diagram to evaluate the impact based on the probability of being right and the cost of being wrong:
Figure 2: Framework for knowing when to deploy an anticipatory service retrieved from Kleber (2017)
For Kleber (2017), "What a designer knows about a user will determine the likelihood of being right". She also makes the reservation that anticipatory design will never be perfect, even with the development of the smartest AI algorithms. Because, anticipation is designed based on routines and from the moment that norm changes, the algorithm is put to the test, and she only anticipates two possible scenarios "Adjust in real time or fail due to lack of contextual understanding." (2017). This framework is the first step to determine if some service is worth to anticipate. Otherwise, we are just using technology just because we can and not because we are proposing to facilitate people's lives. Since in the end, the true nature of technology should not be the opposite of demanding more of our attention. In the end, the cost of being wrong, for the user, will be confusion, lack of trust and anger.
Serendipitous Discovery
Serendipitous and causality discovery will become the next computational challenge. For Zamenopoulos and Alexiou, the concept of anticipation implies circularity, "how can the effect of an action determine the action in advance" (2007, p. 412). The core activities of a designer are the preparation of a solution for a particular future state of a problem, need or goal that may or may not be previously expressed (e.g. industry disruptive products or services). With the introduction of anticipatory design on the equation, the design(-ing) of this ultimate cause has become a paradox in design under autonomous systems.
Currently, the final solution is constructed by its very own design process (Zamenopoulos and Alexiou, 2007). The design process drives the final solution, and there is a causality effect. Anticipatory principles challenge these design principles, where the process defines the solution.
If we introduce the variant of designing for an act in preparation for a particular effect or future state (Zamenopoulos and Alexiou, 2007), the current causality effect between the process and the outcome need to be rethought. Accordingly, this method will introduce constraints to serendipity and causality discovery in user interaction with full-autonomous predictive systems. Also, the Experience Bubble exposes a possible loop under anticipation systems. And it translates into the possibility that the user gets stuck in a loop of returning events, actions, and activities because the algorithm is anticipating the same needs and acts on behalf of the user. Without the notion of serendipity or causality, the system can be trapped in this Experience Bubble because it cannot interpret the meaning behind actions.
Serendipitous and causality discovery will become the next computational challenge. For Zamenopoulos and Alexiou, the concept of anticipation implies circularity, "how can the effect of an action determine the action in advance" (2007, p. 412). The core activities of a designer are the preparation of a solution for a particular future state of a problem, need or goal that may or may not be previously expressed (e.g. industry disruptive products or services). With the introduction of anticipatory design on the equation, the design(-ing) of this ultimate cause has become a paradox in design under autonomous systems.
Currently, the final solution is constructed by its very own design process (Zamenopoulos and Alexiou, 2007). The design process drives the final solution, and there is a causality effect. Anticipatory principles challenge these design principles, where the process defines the solution.
If we introduce the variant of designing for an act in preparation for a particular effect or future state (Zamenopoulos and Alexiou, 2007), the current causality effect between the process and the outcome need to be rethought. Accordingly, this method will introduce constraints to serendipity and causality discovery in user interaction with full-autonomous predictive systems. Also, the Experience Bubble exposes a possible loop under anticipation systems. And it translates into the possibility that the user gets stuck in a loop of returning events, actions, and activities because the algorithm is anticipating the same needs and acts on behalf of the user. Without the notion of serendipity or causality, the system can be trapped in this Experience Bubble because it cannot interpret the meaning behind actions.
Lack of Heuristics
New heuristics under implications such as fairness, accountability, interpretability, and transparency of the service are needed (Riedl, 2019).
New heuristics under implications such as fairness, accountability, interpretability, and transparency of the service are needed (Riedl, 2019).
The current set of design principles from Rams, Nielsen, Norman and Schneiderman are insufficient for automation because principles regarding transparency, control, loops, and privacy are missing. (Van Bodegraven, 2017, p. 435)
Nonetheless, Google has done work on building a design guide, the People and AI Guidebook (PAIR), which establish the first guidelines in the design field for the best practices when designing for AI-driven products or services.
Conclusions
The current state of anticipatory design principles can become dangerous in the sense that it relegates the user decision-making process to a second plane. The first plane is only focused on anticipating users' needs before a specific need even arises in their minds and customizes the content or action accordingly. We understand that the primary goal of the method is to reduce decision fatigue.
However, this ability of an autonomous system to predict user decisions, removing them, in a certain way, from the equation and leaving them entirely out of the decision-making process can become a double-edged sword. because what will become from the human experience if decisions are made for us? For Van Bodegraven (2017), feedback loops should be implemented in the system to allow users to have a say in the interpretation of machine-learning based systems. This strategy will help to decrease the change of inaccuracy as well as rising the trust levels within the service.
The opportunities to reduce users' cognitive overload is very bright and promising in the design field. However, with fully autonomous predictive systems, serendipitous and causality discovery may be removed from the calculation, provoking an Experience Bubble around an individual or even collectively. The design field needs more studies around these matters, especially, more in-depth analysis in the opportunities and risks on what anticipatory design brings to the field of design.
The field of design needs a ground set of principles and heuristics that may help designers design transparent and controllable systems. Otherwise we will not be able to unleash the full potential of anticipatory systems. Engineering may solve the technological challenges, but it lays in the designers' shoulders to fully resolve the user experience inherent to such systems.
The first generations of anticipatory applications have been released. How is time for the design field and the designers' communities to take a step back and figure out how to take these new technologies to the next level.
However, this ability of an autonomous system to predict user decisions, removing them, in a certain way, from the equation and leaving them entirely out of the decision-making process can become a double-edged sword. because what will become from the human experience if decisions are made for us? For Van Bodegraven (2017), feedback loops should be implemented in the system to allow users to have a say in the interpretation of machine-learning based systems. This strategy will help to decrease the change of inaccuracy as well as rising the trust levels within the service.
The opportunities to reduce users' cognitive overload is very bright and promising in the design field. However, with fully autonomous predictive systems, serendipitous and causality discovery may be removed from the calculation, provoking an Experience Bubble around an individual or even collectively. The design field needs more studies around these matters, especially, more in-depth analysis in the opportunities and risks on what anticipatory design brings to the field of design.
The field of design needs a ground set of principles and heuristics that may help designers design transparent and controllable systems. Otherwise we will not be able to unleash the full potential of anticipatory systems. Engineering may solve the technological challenges, but it lays in the designers' shoulders to fully resolve the user experience inherent to such systems.
The first generations of anticipatory applications have been released. How is time for the design field and the designers' communities to take a step back and figure out how to take these new technologies to the next level.
References
Aaron Shapiro. 2015. The Next Big Thing In Design? Less Choice. Retrieved April 22, 2018 from https://www.fastcompany.com/3045039/the-next-big-t...
Alan Cooper. 1999. Inmates Are Running the Asylum, The: Why High-Tech Products Drive Us Crazy and How to Restore the Sanity. Sams Publishing, United States. https://doi.org/10.1007/978-3-322-99786-9_1
Anne Quito. 2015. The next design trend is one that eliminates all choices.
Barry Schwartz. 2004. The Paradox of Choice, Why More Is Less. Retrieved from https://www.researchgate.net/publication/200552951_The_Paradox_of_Choice_Why_More_Is_Less
Christopher J. Anderson. 2003. The psychology of doing nothing: Forms of decision avoidance result from reason and emotion. Psychol. Bull. 129, 1 (2003), 139–167. https://doi.org/10.1037/0033-2909.129.1.139
Colin M. Gray, Yubo Kou, Bryan Battles, Joseph Hoggatt, and Austin L. Toombs. 2018. The Dark (Patterns) Side of UX Design. Proc. 2018 CHI Conf. Hum. Factors Comput. Syst. - CHI '18 February (2018), 1–14. https://doi.org/10.1145/3173574.3174108
Edmund Husserl and John B. Translator Brough. 1991. On the phenomenology of the consciousness of internal time (1893-1917). https://doi.org/10.1017/CBO9781107415324.004
Eli Pariser. 2011. The Filter Bubble: What the Internet Is Hiding from You. https://doi.org/10.3139/9783446431164
Fabien Girardin and Neal Lathia. 2017. When User Experience Designers Partner with Data Scientists. In Aaai 2017 Spring Symposia, 376–381.
Graham Dove, Kim Halskov, Jodi Forlizzi, and John Zimmerman. 2017. UX Design Innovation. In Proceedings of the 2017 CHI Conference on Human Factors in Computing Systems - CHI '17, 278–288. https://doi.org/10.1145/3025453.3025739
Golden Krishna. 2015. The Best Interface is No Interface, the simple path to brilliant technology. Pearson Education (US), United States.
Jason A. Clark. 2016. Anticipatory Design: Improving Search UX using Query Analysis and Machine Cues. J. Libr. User Exp. 1, 4 (2016).
https://doi.org/10.3998/weave.12535642.0001.402
Joël Van Bodegraven. 2017. How Anticipatory Design Will Challenge Our Relationship with Technology A Future Without Choice. In The AAAI 2017 Spring Symposium on Designing the User Experience of Machine Learning Systems, 435–438.
Mark O. Riedl. 2019. Human-Centered Artificial Intelligence and Machine Learning. Hum. Behav. Emerg. Technol. Volume 1, Issue 1 (2019), 1–8. Retrieved from http://arxiv.org/abs/1901.11184
Paul Daugherty and James Wilson. 2018. Human + Machine: Reimagining Work in the Age of AI. Harvard Business Review Press, United States.
Philip Van Allen. 2017. Reimagining the Goals and Methods of UX for ML/AI A new context requires new approaches. In The AAAI 2017 Spring Symposium on Designing the User Experience of Machine Learning Systems, 431– 434.
Roberto Poli. 2009. The many aspects of anticipation. Foresight 12, 3 (2009).
https://doi.org/10.3905/jpm.1978.408666
Sarah Doody. 2018. The Balance Between Anticipation & Automation in Design
Sophie Kleber. 2017. How to Get Anticipatory Design Right. Retrieved from https://www.hugeinc.com/articles/how-to-get-antici...
Qian Yang, Nikola Banovic, and John Zimmerman. 2018. Mapping Machine Learning Advances from HCI Research to Reveal Starting Places for Design Innovation. In Proceedings of the 2018 CHI Conference on Human Factors in Computing Systems - CHI '18, 1–11. https://doi.org/10.1145/3173574.3173704
Theodore Zamenopoulos and Katerina Alexiou. 2007. Towards an anticipatory view of design. Des. Stud. 28, 4 (2007), 411–436. https://doi.org/10.1016/j.destud.2007.04.001
Zhiyuan Chen and Bing Liu. 2016. Lifelong Machine Learning. Synth. Lect. Artif. Intell. Mach. Learn. 10, 3 (2016), 1–145. https://doi.org/10.2200/S00737ED1V01Y201610AIM033
Alan Cooper. 1999. Inmates Are Running the Asylum, The: Why High-Tech Products Drive Us Crazy and How to Restore the Sanity. Sams Publishing, United States. https://doi.org/10.1007/978-3-322-99786-9_1
Anne Quito. 2015. The next design trend is one that eliminates all choices.
Barry Schwartz. 2004. The Paradox of Choice, Why More Is Less. Retrieved from https://www.researchgate.net/publication/200552951_The_Paradox_of_Choice_Why_More_Is_Less
Christopher J. Anderson. 2003. The psychology of doing nothing: Forms of decision avoidance result from reason and emotion. Psychol. Bull. 129, 1 (2003), 139–167. https://doi.org/10.1037/0033-2909.129.1.139
Colin M. Gray, Yubo Kou, Bryan Battles, Joseph Hoggatt, and Austin L. Toombs. 2018. The Dark (Patterns) Side of UX Design. Proc. 2018 CHI Conf. Hum. Factors Comput. Syst. - CHI '18 February (2018), 1–14. https://doi.org/10.1145/3173574.3174108
Edmund Husserl and John B. Translator Brough. 1991. On the phenomenology of the consciousness of internal time (1893-1917). https://doi.org/10.1017/CBO9781107415324.004
Eli Pariser. 2011. The Filter Bubble: What the Internet Is Hiding from You. https://doi.org/10.3139/9783446431164
Fabien Girardin and Neal Lathia. 2017. When User Experience Designers Partner with Data Scientists. In Aaai 2017 Spring Symposia, 376–381.
Graham Dove, Kim Halskov, Jodi Forlizzi, and John Zimmerman. 2017. UX Design Innovation. In Proceedings of the 2017 CHI Conference on Human Factors in Computing Systems - CHI '17, 278–288. https://doi.org/10.1145/3025453.3025739
Golden Krishna. 2015. The Best Interface is No Interface, the simple path to brilliant technology. Pearson Education (US), United States.
Jason A. Clark. 2016. Anticipatory Design: Improving Search UX using Query Analysis and Machine Cues. J. Libr. User Exp. 1, 4 (2016).
https://doi.org/10.3998/weave.12535642.0001.402
Joël Van Bodegraven. 2017. How Anticipatory Design Will Challenge Our Relationship with Technology A Future Without Choice. In The AAAI 2017 Spring Symposium on Designing the User Experience of Machine Learning Systems, 435–438.
Mark O. Riedl. 2019. Human-Centered Artificial Intelligence and Machine Learning. Hum. Behav. Emerg. Technol. Volume 1, Issue 1 (2019), 1–8. Retrieved from http://arxiv.org/abs/1901.11184
Paul Daugherty and James Wilson. 2018. Human + Machine: Reimagining Work in the Age of AI. Harvard Business Review Press, United States.
Philip Van Allen. 2017. Reimagining the Goals and Methods of UX for ML/AI A new context requires new approaches. In The AAAI 2017 Spring Symposium on Designing the User Experience of Machine Learning Systems, 431– 434.
Roberto Poli. 2009. The many aspects of anticipation. Foresight 12, 3 (2009).
https://doi.org/10.3905/jpm.1978.408666
Sarah Doody. 2018. The Balance Between Anticipation & Automation in Design
Sophie Kleber. 2017. How to Get Anticipatory Design Right. Retrieved from https://www.hugeinc.com/articles/how-to-get-antici...
Qian Yang, Nikola Banovic, and John Zimmerman. 2018. Mapping Machine Learning Advances from HCI Research to Reveal Starting Places for Design Innovation. In Proceedings of the 2018 CHI Conference on Human Factors in Computing Systems - CHI '18, 1–11. https://doi.org/10.1145/3173574.3173704
Theodore Zamenopoulos and Katerina Alexiou. 2007. Towards an anticipatory view of design. Des. Stud. 28, 4 (2007), 411–436. https://doi.org/10.1016/j.destud.2007.04.001
Zhiyuan Chen and Bing Liu. 2016. Lifelong Machine Learning. Synth. Lect. Artif. Intell. Mach. Learn. 10, 3 (2016), 1–145. https://doi.org/10.2200/S00737ED1V01Y201610AIM033
Did you like this article?
