Line of Adaptivity in HCI and HRI: user experience similarities and differences back

Do you think human-computer interaction (HCI) is a complex topic? Then how about human-robot interaction (HRI)? Computer is just a machine, or at least looks like one and acts like one. However robot does not look just like any machine – anthropoid robots are designed to resemble humans and mimic their behavior. For example a research group at Carnegie Mellon University is designing robotic products and services to be adaptable to people’s changing behavior over repeated interactions. But how would people react once faced with a need to interact with such creatures?

This original article is dealing with enriched Service Blueprint (SB). SB is a tool for service design which we will cover in the nearest future, however this time we focus on implications of repeated service encounters. Our initial idea was to simply present you the model, but then it appeared that the same model has many parallels to human-computer interaction as well. So the case descriptions under each model sections outline some ideas about parallels and differences between HCI and HRI.

Project background


In spite of the fact that in developed countries people eat snacks at least once a day, very few efforts are made to assist and improve such snacking practices. Thus, CMU “Project on People and Robots” researchers are developing a cute robot – called Snackbot – which (who?) delivers snacks to people. Snackbot is a mobile robot, about the size of a very small human, that rolls around on wheels to deliver snacks to students, faculty, and office workers at Carnegie Mellon University. While snack ordering is done via website specifying desired snacks, delivery location and time, when delivering snacks Snackbot interacts with people in a social way, uses natural human language, sound and head/body movements.


As the team admits, the challenge in designing the adaptive robotic service is finding answers to questions on how to:

  • make service flexible (adaptable) to changes of human behavior over time
  • incorporate change of human behavior into service provision
  • maintain user’s interest in service over time

The task is especially complicated, given that humans will have direct contact with robot, not a human being. Based on research on service design, adoption of technology products, and user experience design the authors come up with the idea of enriching Service Blueprint with a Line of Adaptivity . Line of Adaptivity successful integrates 4 time related factors which allow overcoming the design challenges. The four factors are: orientation, incorporation, streamlining, and personalization. Orientation and incorporation refer to changes in people, while streamlining and personalization relate to changes in the components of service.
Line of Adaptivity


At the initial stage of service encounter authors claim sense-making to be of utmost importance in understanding how the service works. It encompasses understanding functional aspects of service, evaluation of their utility and desirability, and formation of initial attitudes towards the service. The more novel the offering (or technology), the greater sense-making activities take place.

Case of Robot service delivery
First, using a two-way interaction between users and the product and service, robot will use speech and visual recognition to capture human behaviors and respond to interaction breakdowns. In this type of interaction robot is programmed to actively evaluate the state of human sense-making about the service and bring back the user “on-track” if necessary.
Second, incorporation of human social cues into robot design will allow people to draw on their experience with familiar service. Examples of such cues are humanoid shapes in robot design (robot looks a bit like human), speech interaction sequence similar to that of a human vendor, etc.
Case of Internet service delivery
We can see strong parallels with Internet based service delivery here. Any new webpage or web application requires sense-making to some extent. Things like design patterns are used to minimize the need of sense-making through interface patterns which are familiar to users from other experiences. However, proactive tracking whether user understands the service flow correctly (and related correction measures) is largely not present in Internet solutions. To paraphrase – also in theory HCI has a well-known concept of mental models, active checking if the sense user makes corresponds to developer’s intentions is not widely present. It’s mostly all about guidance through tips, hints and error messages, but the actual decision whether you are on the right track is still left to the user.



At this stage users begin to integrate service into their daily activities, building trust and emotional attachment. Sense-making becomes peripheral, taking place only when aspects of service do not fit into previous experience.

There are three so-called cultural models helping to understand how people respond to services: relational, oppositional and utilitarian. If service uses only one setup, depending on the prevailing model used, different customers would have either pleasant or dissonant experience with the same service.

  • Relational model is applied by people who desire and value emotional ties with a service provider.
  • Oppositional model is applied by people who perceive themselves as vulnerable, weak players in the consumer-provider relationship and easily take an aggressive stance toward the service provider.
  • Utilitarian model is applied by people who rationally weight service benefits against costs.
Case of Robot service delivery
Understanding which cultural model customer applies is done via recording and tracking down how people talk to a robot. Furthermore, to facilitate better service depending on customer’s cultural model, they can choose on the website what type of interaction they desire during service provision.For relational cultural model, Snackbot mimics human vendor interaction and applies interpersonal relational strategies:

  • follows the rule of reciprocity in giving a free snack on some special occasion (thus expecting to create feeling of thankfulness and indebtedness towards the robot)
  • refers to past dialogues with user
  • use self-disclosure strategies to create feeling of closeness

For utilitarian model, Snackbot follows a more machine-like interaction without trying to engage users in social conversation with goal of delivering an efficient, minimal transaction.

Case of Internet service delivery
While the three models could probably be observed in Internet service delivery as well, based on our user testing experience a fourth model comes to mind in the context of computer use. That is something we would call Inferiority model –many users (especially less experienced ones) are quick to jump to conclusions that it’s their fault if software or website does not work (I am not skilled enough, I don’t know how to use it), thus their whole user experience is largely about blaming themselves for things they are not able to achieve.To some extent this could be a model on the other extreme from Oppositional model. While inexperienced internet users are likely to blame themselves for unsuccessful user experience, experienced users (especially the ones with awareness of usability requirements), are quick to take aggressive position toward the system and blame it/service provider for poor performance and difficult interface.



Once user’s routines and preferences are learned, some of the touch-points may become unnecessary. Thus, to ensure positive user experience, step (procedure) elimination should be supported (idea similar to shortcuts in software).

Case of Robot service delivery
Snackbot will be able to combine or automate some steps in the service journey for expert users. Speech-based instructions on how to complete the transaction with robot might become unnecessary. Instead, simple sound indicating actions might be used (for arrival or approval). In case users exhibit certain patterns over time (say ordering cookies every Friday), service automation can take place (one click confirmation).
Case of Internet service delivery
This part of the Line of Adaptivity concept can probably be most closely applied to both, robot and internet service delivery. Internet systems also have a concept of providing more instructions to novice users and allowing process shortcuts for more experienced ones. Since the use experience can be estimated easily based on visit frequency and time spent, automated streamlining is easy to implement in Internet systems.



Services can be personalized to better fit the needs of users, once their patterns of service purchase and use are learned.

Case of Robot service delivery
Personalization is mostly achieved via tracking preferences and customizing responses. For example, for customers preferring healthy lifestyle robot offers healthy snacks. Via engaging into conversations with users following relational cultural model, robot can learn of special occasions and deliver snacks on them. But most importantly, service can attempt to understand diverse motivation on why people order snacks and customize the service in response (for example, for people who use service to enjoy social snack break to socialize with colleagues or friends , the service can facilitate coordination of more people at the same time. Let’s say, robot will inform that your colleagues in other room also have a snack break now, so maybe you want to join them).
Case of Internet service delivery
Personalization is a buzz word in Internet services already for a while. Everyone is talking that information and services should be personalized so that they would be available to the user in the right context and format at the right time.However, the interesting difference compared to the robot service design is that Internet services still rely mostly on user’s own actions to personalize things for himself (be that information or visual settings or something else).

Why aren’t internet service designers relying on machine learning technologies to learn user preferences and personalize services automatically?


How about conclusions? Well, in relation to the Line of Adaptivity concept we find it easily applicable and very useful to human-computer interaction field. Also while reading this and other HRI research articles, we, ourselves primarily being software user experience experts, cannot stop wandering why IT field is not as concerned with automated adaptability of services provided through Internet portals/software interfaces as robotics field.

More readings

Original article: Min Kyung Lee, Jodi Forlizzi. Designing Adaptive Robotic Services (PDF)

More about Snackbot design: The Snackbot: Documenting the design of a Robot for Long-term Human-Robot Interaction (PDF)

Share this..Share on FacebookTweet about this on TwitterShare on Google+Share on LinkedInPin on PinterestEmail this to someone