ACI@BHCI 2015

On the 13^th of July, British Human Computer Interaction 2015 (BHCI’15) hosted a workshop on Animal Computer Interaction (ACI) coined ACI@BHCI’15, in Lincoln, England. This workshop had talks from previously well known ACI researchers as well as fresh researchers within the field. The workshops hosted two activities, ‘Bad Design in ACI’ by Ilyena Hirskyj-Douglas and Charlotte Robinson and ‘Design Fiction in ACI’ by Ben Kirkman. This blog post will map the questions raised during this conference. This blog will also talk about the workshop activities and outcomes to give those who were unable to attend an insight into the day.

PAPERS, PRESENTATIONS AND QUESTIONS

Ilyena Hirskyj-Douglas, Janet C Read & Brendan Cassidy. Doggy Ladder of Participation. (University of Central Lancashire, UK)

Link to Paper: Doggy Ladder of Participation

Ilyena presented her theory paper on how to include dogs within design through participation. This can be measured using a ladder model based upon Harts Ladder of Youth Participation she had coined ‘Doggy Ladder of Participation: DLOP’. The first question raised about this model was, does the author see this model working on other animals. Ilyena felt this model would work on other animals but it would have to be slightly adjusted to allow for their individual communication methods. A comment was also made by a researcher who felt that the ladder would be straight but curved instead. Another researcher was concerned that a dog would not be able to grasp the whole system. While Ilyena believed a dog would never be able to understand the whole system she did believe a dog would be able to understand the input causes a reaction understanding how the system works and not necessarily the internal coding. The final discussion around this paper was based around whether the concept of no training could exist in a system, or if self-learning a system was a form of training. One of the key notions behind the DLOP is the idea of not training a dog to use a system but letting the dog use the system ordinarily to get unbiased results. The author sees a distinction from this method to self exploration where the dog would explore the system to find out its activity.  The difference between this self exploration notion and training was discussed, as researchers in the animal department at Lincoln argued that self exploration was a form of training. Ilyena on the other hand felt that training and self exploration were two different concepts, lead by two different species although could be closely interlinked.

Felipe Raglianti. Position Paragraph. (University of Lancaster, UK)

Link to paper: Felipe Raglianti Position Paragraph

Felipe presented his position paragraph on a system informed by ACI that could fund relationships between non-human animals and humans in “becoming with each other, to become human through technology”. This included the relationship between humans and dogs being a social relationship. Similarly to Ilyena, Felipe got asked questions around participation and training. He felt training could be a vehicle of participation, where there is a way of training but the dog is also training the human in a two way street analysis, in the same way Ilyena felt there was an ongoing conversation machines need to be involved with. Felipes’ talk was primarily around interface design, so naturally there were questions around interfaces for dogs. The University of Lincoln Animal Department, asked if he would change the interface for the dog (such as screen rate etc.) which Felipe said he would as the interface matters so much for dogs. This opened up to a discussion on how animals see screens and how should you design interfaces. Felipe felt designing for species is a game of approximation. This is similar to Charlotte Robinson work with diabetic dogs where she had to try different textures for a handle to suit the dog. Felipe thought you could test the screen in a purely behavioural way, involving elements of user centered design.

Steve North, Amanda Rosher, Carol Hall & Clara Mancini. HABITAT a Horse Automated Behaviour Identification Tool – A position Paper. (University of Nottingham, UK: Nottingham Trent University: Open University).

Link to paper: HABITAT

Link to Facebook: HABITAT Facebook

Link to workshop presentation: Youtube Talk

Steve North presented his position paper on HABITAT a proposed system that has the ability to automated analysis and recognition of horse behaviours both in solidarity and within a group. He is currently crowd sourcing video clips of horses, leading to the next stage of developing the tracking technology. He ideally would like real-time recognition but initially is aiming for pre recorded recognition through the video datasets. He felt this tool would be both an ACI system and also a support tool to aid the ACI process. Unlike Felipe who held a very ethnographic approach, Steve believed in taking from HCI to aid ACI and would argue that ACI includes animal species studies. Steves’ ideology was that HABITAT would work both in the field and complex habitats, tracking both individual animals and herds. Ilyena, having built a dog face tracker herself, asked if he planned to build a skeletal framework for the horse or track on features. Steve said he had not yet gotten to that stage within HABITAT where he planned on bringing in a visual computer specialist. He did though plan to do the tracking upon outlines and characteristics of movement: ‘ground truth images’. There was a lady from Lincoln’s Animal Department that commented that she tracked with fish and a huge amount of videos go into making a dataset. The last comment made about HABITAT was around privacy: would the horse act different if he/she knew he/she was being filmed? Steve said he faced this dilemma at the moment as all the video datasets were gathered from the public where he suspected the horses knew about the camera’s presence but ideally a camera would be unobtrusive to allow for natural behaviour.

Germain Lemasson, Dominique Duhaut & Sylvie Pesty. Dog: Can You Feel It? (Lab-STICC, France: LIG, France).

Link: Can you feel it?

Germain attended ACI@BHCI to present his work on calling back dogs for wheelchair users using a haptic harness. He is currently working on a project to give remote commands to a working dog based upon the work of Miller. This system uses a Smartphone’s GPS to recall dogs automatically. The haptic harness creates rubbing that is natural and he thinks the dogs enjoying the system but ultimately he does not know (Germain did mention that he had tried it himself!). There was a suggestion that similar to Adrian Cheoks huggable chicken jacket; there could be two rooms (one with and one without) to see if the dog enjoyed the jacket. This was further talked about with a touch screen to enable the dog to turn on/off the jacket, or GPS boundaries to allow the jacket only to be on when in certain locations. Questions were also raised here, as Adrian Cheok suggested the jacket relaxes the chickens being stroked, did Germains jacket relax the dog? Germain did say though that he wanted to start using particularly design to see how the dog feels using the jacket. Germain found the dog stopped listening to the system when no reward was available but did not want to put food within the harness in fear that the dog would chew the electronics. There was also a question raised of if the dog could feel the tactile over their fur, which Germain said the jacket was rather strong so it could get around that barrier. A suggestion was made that perhaps the dog could learn just from vocal commands without a reward so make the stroking of the jacket the reward, maybe combining with a toy. However some dogs do not like to be touched so this system would not work for them. Another point made around Germains work by Clara Mancini was that sometimes stimuli are influenced by the way the animal feels about the meaning of the interaction with the whole experience coming from a bigger experience. This experience might not be associated with a certain feeling so isn’t yet within the spectrum. Germain had not yet explored this, with at first thinking that the owner could tell him what the dog was interpreting, but he feels even when you train a dog to learn behaviour and vocalisations from humans it is very important to do it within the context but still the dog might not know how to react. Another suggestion was made to maybe include smell so wouldn’t have a problem with the food incentive. This idea came from the smell of cinnamon used to stop dogs so maybe a good smell could be used to entice. This is further backed up by dogs using pheromones, so a suggestion was made that maybe pheromones rather than a food smell could be implemented. This idea was then explored in the context of meaning, where the smell could be the owners smell or a stroke given by the jacket that is similar to the owner. Germain had thought of this idea and stated that he debated putting the owners’ cloths on the harness. During this conversation a researcher did point out that electric collars do vibrate, so at what point does the vibration become a reward?  Germain had experimented with vibrating collars but stated that the dogs do not like it as the neck is a predatory area with little context. Finally Clara Mancini commented about the similarity to Temples Grandins’ work, with a squeeze machine, where it was seen to have a calming effect so maybe by contracting the device the dog might respond differently. Germain said he had started to go down this route but there were technical issues.

Anna Zamansky, Sofya Baskin & Sharon Anavi-Goffer. Digital Game Design for Canines: Getting to Know Your User? (University of Haifa, Israel: Ariel University, Israel).

Link: Digital Game Design for Canines

The last talk of the day was given by Anna Zamansky who began her presentation by acknowledging that dog users are very central to ACI and that animal play is an activity that is self rewarding and relaxing with different types of play (object orientated and social play). She got interested within this field to improve the welfare of dogs, their physical activity and provide insights into the nature of human play. He work is held within two phases: the first phase being the cataloguing of behaviours of a small-medium dogs playing with a tablet from YouTube, and the second a 2 dog experiment (Poodle, 1 year old: Jack Russell, 5 years old) of their reaction to an interactive digital game that responded to their touch. The second phase was used to validate the ethograms from the first stage within a controlled context. She found that the two participants exhibited different behaviours, the Poodle liking the tablet and the Jack Russell exhibiting negative behaviours. Questions at this stage were asked if the human was paying attention to the video as it is well known within ACI that this can affect the outcome of the experiment. She believed this showed two different models of stimulation, enjoyment and over stimulation. She planned on extending this study to various participants to see if personality, breed etc held an impact. The first question raised was about how much influence the owner had over the dog. Anna said that both of the dogs where the researchers own dogs, and they did not interact with the owner or the owner with the tablet during the study. Felipe was intrigued by the Jack Russell discovering the screen as we do not know what the screen was for the dog as the dog was still looking for the mouse within the video once the screen had gone on the floor. We can never know how a dog perceived a screen but Sofya was expecting to see more game movement but this did not happen. The next question raised was around social interaction not taking place within the study and maybe there are different markers with animals and machines in social play. Anna mentioned that when dogs play with robots they have seen to be different in play than with other dogs, or humans, so maybe it could be the context and markers that change (i.e. the robot is something physical but a tablet is flat). This could also be down to the individual and it would be interesting to see the range of interaction and where dogs cross the object-social line within play. Anna agreed that there was a lot more work to do within this area and with tablets.

Workshop Activity 

The afternoons activities were centered on designing bad ACI and Design Fiction. For this activity we were split into groups to discuss the topics.

Activity: Bad Design

For this activity three scenarios were presented to the workshop where as researchers we had to come up with solutions to these problems that would be bad. This is unusual in ACI as usually we are looking for the ‘good’ design. This activity got the researchers to think about what bad design is and thus opposed what is good quality design. The three scenarios given were of

• A cat that refused to use a normal cat flap and wanted an automated system
• A pig that wanted an automated bath
• A parrot that wanted to scare the owner through the use of a jack-in-the-box

Cat Flap

Bad Design

• Mechanically pull on a string
• Meow trigger (meow triggers the dog opening)

Good Design

• Chin tickler door opening. This works by when your kitty comes along a mechanical hand will stroke the cat causing the cat to purr. A computer system will then recognise this (voice recognition) thus sliding the cat-flap open.

Questions Raised

• How do you train a cat to use such a system?
• What is the cat gets an illness and could no longer use the voice recognition?
• What is a cat enjoys the hand stroking and instead uses the system for its alternate use?
• Would the cat prefer instead to have an electronic tag that simply just magically opened the door?

Pig Bath

Here it was seen that a relationship between the input and output was important

Bad Design

• Cords all around the bath that the pig has to pull but the cords only work at a certain time of day on alternate weeks with time delays
• Shower for the bath has an inconsistent water supply
• Little cords so hard to use and pigs are unable to reach the cords
• Texture of the cords not nice to the pig

Good Design

• Controller next to the bath
• Pig puts its nose in the hole and it will trigger a pigozzy (pig jacuzzi)
• Water comes from the bottom of the bath so not to startle the pig
• Consistent water supply
• Feedback given to the farmer to give an indication of how the pigs are using the bath

Parrott jack-in-the-box

Bad Design

• Controller similar to dance-dance revolution with coloured blocks that the parrot has to jump upon in order to work the jack-in-the-box
• The pattern needed to work the jack-in-the-box was constantly changing
• The parrot had to do this in the dark and back within its cadge
• A horrible noise would be given on the wrong answer

Good Design

• The pattern was constant and short
• The parrot could do this in or out the cage and within the light
• There would be no terrible noises made

Summary from Bad Design Activity

• Everyone made assumptions upon the animals which brings forth the need for ACI researchers to work closely with Animal Behaviourists. Often when designers do not know a lot about the animals they project their own views into the animals’ desires.
• The designs that were seen as good do not require training but intrinsically use the ordinary behaviour of the animal
• If the animal is aware that they are interaction with a system does that motivate them to use a system? E.g. Is the pig happy that they are working the bath? Does this empower the animal? This point demonstrates that we are not yet aware of the associations that animals make to comment on this about a system.
• Rather than awareness it is whether the anima wants something which can be shown through their behaviour and as long as this contributes to maintaining good health then this is respecting the autonomy and agent of the animal.
• Maybe through understanding an animal’s needs more this might change our attitude to designing ACI. Unfortunately it is a constant game of approximation where we have to do what we think is best.
• Bad design = inconstant results
• ACI is very entangled within ethical boundaries and in fact what is the boundary of ACI? The kernel at the root of ACI is about making things better for animals but it is about making the processes more efficient. However in order to do this you have to make things better for the user in order to be more proficient.
• Questions were raised on the relationship between ACI and HCI. It was generally felt that HCI is a subset of ACI if we accept the theory of evolution: we all came from animals.

ACTIVITY: DESIGN FICTION

Ben Kirkman from the University of Lincoln ran the Design Fiction activity where the researchers in ACI were challenged to come up with futuristic designs that would not be beneficial to animals. This idea came from Ben’s previous research into a collar for dogs that allows, through emoticons, the dog to express how it feels to the owner. This collar was a design fiction, and he made a website that made this fiction seem real and asked owners to comment upon the product. He found that owners would often trust the collar over the vet and ignore a vet’s advice if the collar said the dog was happy anyways. This point is very prominent among ACI with the upsurge of technologies for pets that do not necessarily improve the pets’ welfare.  This activity was split into three sections

• Privacy
• Wearable Computing
• IoD – Internet of Dogs

Privacy

This group discussed privacy within ACI:

Within ACI privacy is seen as something that any animal can ask for with rejection of a system – but it is not simply about rejection but also about the research methods being ethical creating a double definition on both parties leading to an open ended agreement. It was discussed that you see a constant rejection you stop a study but there needs to be more designs in technology that allows an animal to assess the situation and negotiate a stop to a study. This raises the question of how do you figure out a licence agreement between that parties that can be rejected by the animals.

Wearable Computing

This group came up with some Design Fiction wearable technologies for animals:

• Augmented dog bark to allow a small dog to scare off a larger dog
• Kitty and Dog VR Glasses ‘Dogulus Rift’
• Bracelet or collar that sprayed the animal with glitter perfume
• App that allows you to point your Smartphone at a dog to see if it is ‘aggressive/bad’
• Wearable jackets on ants that allows them to be controlled by a computer and playable by a human ‘SIMS Ants’
• Shoes that allow an animal to feel like they are walking on a different textures e.g. grass/sand

IoD Internet of Dogs

This group came up with some Design Fiction about the Internet of Dogs (IoD)

• Doggy Skype
• Treadmills that the dog has to run to get food
• Internet of Dogs Facebook ‘Dogbook’
• Dog Uber – Rent a dog

Summary & Thanks

Overall the day was very enjoyable with lots of interesting conversations happening. I would like to take this opportunity to thank all of those who attended, either physically or via Skype for making ACI@BHCI a wonderful event.