Attending CHI 2017

During May 6th to 11th, I, together with my group mates, attended the ACM SIGCHI 2017 conference in Denver, USA.

The ACM SIGCHI conference, namely Special Interest Group on Computer-Human Interaction, is a premier international conference on Human Computer Interaction. This year’s conference, CHI 2017, was held at Colorado Convention Center in Denver. Our group arrived in Denver one day earlier to adjust to the time difference as well as the high altitudes in Denver. After settling down, we started to prepare for the next day’s presentations and schedule the session attendances during the conference period.

Presenting Paper

Our paper, ``Sensing and Handling Engagement Dynamics Involving Peripheral Computing Devices in Human Robot Interaction”, was invited to present at session Robot at Home & Work on the first day of the conference. This paper summarizes a part of work we had done for the WeChat project. Specifically, we described in this paper the models implemented for robots to detect participants’ engagement shifts during their interactions with robots while they are also using peripheral computing devices. We also proposed in detail the development of associated strategies for robots to make appropriate responses.

Serving as the first author, I was delegated to present our work in the conference. It was my first time to be the first author for a published paper, and also my first time to present our work in an international conference. I was quite nervous before the presentation, partly because I did not even present well during the rehearsal in our weekly group meetings. Even for the last rehearsal in Denver, my supervisor was still not satisfied as her straight faces showed. So, in the presentation morning, I found a corner to practice the presentation again and again until I could memorized all the scripts written for it (scripting is not the good way to present, but for a non-native English speaker, it is effective to avert any possible disasters). I went to the designated session room half an hour earlier, and found nobody there. Fortunately, it gave me the change to experience what it might feel like when standing on the stage. I walked onto the stage then and stood there for a while to calm down. Then people started to come in and I walked down quickly to find a seat on the first row of chairs. Almost instantly, the room was filled with tens of audiences. Without much waiting, the session host walked on the stage and announced the opening of our session. I knew it was time to present. At the very beginning, I was nervous, yet pretending hard to be confident, to walk again onto the stage. After a while, as I started to introduce my self and my work, I calmed down a little and managed to follow the underlying logics of my slides. Fifteen minutes passed away so quickly that, after I took two questions from the audience, the presentation was over. I felt greatly relieved. I walked down the stage and meanwhile I noticed my supervisor was smiling at me. Apparently, my presentation was not a disaster.

Attending Sessions

After the presentation, there was almost little pressure to bear on me. So I checked all the sessions and tried to spot any one related to my current research. I picked several sessions on robots and vision-based sensing/tracking to ensure that no time slots conflict with each other.

One interesting presentation I attended was about how to get rid of the influences of commercial advertisements. The presenter discussed that, in nowadays, we were all inevitably surrounded by different ads, which could potentially affect our daily decisions. However, there had been no comprehensive studies on techniques to dismiss such influences. The presenter, thus, implemented a system, called Anti-Influence Engine, to help people escape from the diabolical machine of pervasive advertising. This idea subverted my opinions on the advertisements and the beliefs on the people’s decisions. Before that, I had been thinking that each people, as an independent individual, is supposed to have his/her own belief, which could not be easily shaped by their surroundings, let alone those ads. But the truth is such belief could not even stand the test of advertising. And people start to need machines to help hold their beliefs.

Another odd yet novel presentation was about a new technique called EarFieldSensing (EarFS), which collects electric field changes and physical deformations caused by face muscle movements, and exploits them to estimate facial expressions. The authors developed such novel approach to sense humans’ internal status, and successfully achieved high detection precision compared to other electric sensing technology (EMG, CS, EFS etc.). Detecting/Estimating humans’ facial expressions/gestures has been studied for years, and various approaches are proposed to improve the performance. However, until now, there has been no significant progress in this domain. Conventional vision-based or head-mounted-device-based methods appear to lose their momentums and struggle for better performance. Such novel technique, i.e., EarFS etc., provides a new insight to address this detection difficulties.

Joining Panel Discussions

Panels in CHI conference are 80-minute sessions that involve significant interaction with audience members. Usually, a panel session includes a group of experts invited to debate a topic or theme, enact some aspect of their expertise, or reflect on and compare their diverse experiences. In addition, a panel will also involve audiences’ interactions, e.g., through questions and answers, voting or critique of the experts’ presentations or extensive discussion. In this CHI conference, researchers from IBM organized a panel discussion on Human-Agent Collaboration: Can an Agent be a Partner. I was fortunate to attend this session since so many people flocked to the venue that all seats were taken instantly.

At the beginning of the panel, the host gave a brief introduction, and proposed two starting points of the upcoming discussions. These points were the fundamental difference between interaction and collaboration, and the trustworthiness of agents. Experts from different institutes, including IBM Waston Research Center, Microsoft Research, Google Inc. etc., were then invited to voice out their opinions. Timothy Bickmore, from Northeastern University, focused his talk on the humans’ perceptions towards therapeutic alliance agent. He mentioned that, by changing agents’ narrative point of view to 1st person, the patients could perceive agents more as human-like companions. Thus, agents were capable of being a collaborator in the therapy context. However, Sean Andrist, from Microsoft Research, argued that agents could only serve as complements, instead of human duplicates, from a more practical perspective. He spoke that current agents were functioning more as tools and still facing many challenges of being potential collaborators, such as coordinating with humans and managing humans’ expectations. In addition, he pointed out the lack of effective models and representations for collaborative activities in current research. Thomas Erickson, from IBM Waston Reserach, supported Sean’s arguments by his own claims. He showed current agents’ inabilities of being a collaborator from several angles, such as their lack of intelligence of providing significant assistance with most tasks or serving as an assistance, their lack of social abilities, and peoples’ unwillingness of having partners. Rachel K.E. Bellamy, also from IBM Waston Research, gave another insights by proposing that more notices should be taken of humans, who are now prone to habitually behave like robots. She argued that, instead of developing intelligently collaborative agents, we human beings should received more trainings before we could really work with them. All there insights inspired me, a junior robotic student, to dig deep to the study of Human Robot Interaction.

Participating Other Activities

Apart from those formal sessions, there were also many other activities organized by sponsors and local universities. Our group participated in CHI parties held by UCL Universities, University of Washington, Google Inc etc. Many companies and research institutes mounted additional exhibitions to showcases their research outcomes or edge-cutting products.

Thoughts and Insights

During CHI 2017, we attended many sessions, exhibitions, and university parties. We also met with other researchers or designers. All of these experiences have impressed me and will motivate me for a long time. Conference offers one channel for researchers to communicate with each other. Essentially, academic conference is research-oriented: researchers from all around the world discussing and presenting their work. In the meanwhile, it is also big social party: people trying to know each other and also pitching their new thoughts and ideas. For a non-native English speaker, the language might be the main barrier for us to thoroughly enjoy this big party (talking freely in English in such a big social environment is difficult for me). But as a junior CHI student, I believe there is always too much to learn, not only in research abilities but also in social skills.


The author gratefully acknowledges financial support from UGC Research Travel Grant (2016-17). The awarded project number is RTG16/17.EG316, and the awarded funds are HKD 11,875.