IEEE Robotics & Automation Magazine - June 2019 - 72

also summarize current findings of human-robot interaction
(HRI) research in the field of transparency to show how
transparency works in practice. Finally, we conclude by proposing a checklist for designers outlining a step-by-step guide
that may help robot developers implement the GDPR's transparency requirement. Future work will address the application to individual use cases.
Different Expectations of Transparency
Transparency usually refers to things and concepts that are
easy to perceive or detect. In the context of computing, however, it counterintuitively refers to processes or interfaces that
function without a user being aware of them [21]. This latter
understanding of transparency contrasts with the GDPR,
which demands transparency for information technologies
in the sense of making data processing explicit to the user.
Such standards could entail a barrier to the deployment of
robotic systems that process personal data in Europe or of
European citizens.
The GDPR is intended to be a technology-neutral piece of
legislation, meaning that no specific technology should be the
target of the law. Instead, it should apply to all possible technologies at large. The GDPR's strength lies in providing general
legal requirements across technologies. However, its lack of
recognition for specific technologies and context factors risks
neglecting crucial elements in protecting users' data-related
rights. This challenge also arises in the determination of the
requirements of transparency, which will need to be molded
according to the characteristics of the technology-in this
case, a robot. Moreover, complex technologies, such as artificial intelligence (AI) and robotic systems, raise particular challenges not only because of their information processing nature
and contexts of use but also because of the multitude of stakeholders potentially affected by the transparency requirements.

Distinct from many other information technologies, the
end user in an HRI context is not the only user who is engaging with the system. The broader context of robot deployment, involving different roles and responsibilities among
various stakeholders, demands a comprehensive understanding of transparency (see https://standards.ieee.org/
project/7001.html). A typical robot ecosystem, in the healthcare sector for instance, would involve the health-care management team in the organization who initially decided to
deploy the robot, the health-care staff who implement the
robot in therapies or daily care, the family members who
make decisions about their relative's engagement with the
robot, the end user, the robot's developer, and the infrastructure providers [8], [16].
The investigation by Weller [23] into the roles and types of
transparency in the context of human intelligibility of robotic
systems explores this issue (Table 1). However, further differentiation among the various stakeholders in the field of assistive robots awaits development.
Different stakeholders have different roles, information
needs, background knowledge, and abilities. Accordingly, the
transparency requirement needs to be tailored to the types of
users (in light of their roles, responsibilities, and interests)
and their level of capacity and vulnerability. Assistive robotics, as a field, frequently targets vulnerable users, such as
elderly individuals with dementia or children with autism.
Even for nonvulnerable users, understanding the information provided about a specific robot is a nontrivial task. The
inclusion of vulnerable users creates particular challenges for
transparency because strategies regarding transparency that
work for nonvulnerable users may not work for vulnerable
populations. Evidence about effective strategies that do justice to the specific needs of vulnerable populations still needs
to be gathered.

Table 1. Transparency expectations for different stakeholders [23].
Transparency in the Context of Robotics and AI
For...

Transparency serves to...

Developers

Understand whether their system is working properly, to identify and remove errors from the system or
improve it

Users

Provide a sense for what the system is doing and why, to enable intelligibility of future unpredicted action circumstances and build a sense of trust in the technology
Understand why one particular decision was reached
Allow a check that the system worked appropriately
Enable meaningful challenges (e.g., to credit approval or criminal sentencing)

Society broadly

Understand and become comfortable with the strengths and limitations of the system
Overcome a reasonable fear of the unknown

Experts/Regulators

Provide the ability to audit a prediction or decision trail in detail, particularly (un)intended harmful actions, e.g., a crash by an autonomous car

Deployers

Make users feel comfortable with a prediction or decision, so that they keep using the system
Lead a user into some action or behavior, e.g., Amazon might recommend a product while providing an
explanation that the user then clicks through to make a purchase

IEEE ROBOTICS & AUTOMATION MAGAZINE

JUNE 2019

https://standards.ieee.org/project/7001.html https://standards.ieee.org/project/7001.html

IEEE Robotics & Automation Magazine - June 2019

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - June 2019