Measures of Central Tendency: Mean, Median and Mode
Technology and Health Promotion: Implications for Clinical Psychology Practice
1. New Technologies for Health Promotion
Implications for Clinical Practice
Dr. Eric Hekler
Assistant Prof, Nutrition
& Health Promotion
Arizona State University
ehekler@asu.edu
@ehekler
www.designinghealth.org
Presentation for University of Arizona
Psychology and Technology:
Engaging your Clients Workshop
Flickr-RansomTech
13. The process of creation is changing
• Expert-sourced
• Crowd-sourced
o Content generated by
“experts”
o Content generated
by “crowd”
o Information evaluated
by experts
o Information
evaluated by crowd
o Rigorous but slow
o Fast but inaccurate?
o www.britannica.com
o www.wikipedia.org
Flickr – miyagisan
15. A DIY Self-Experimentation
Toolkit for Behavior Change
Win Burleson
Eric Hekler
Winslow Burleson
Jisoo Lee
Arizona State University
Bob Evans
Google
18. What you need to know…
• How our interventions work is changing
– Other ways to provide interventions
– Multi-level/multi-component interventions
• Psychologist’s role is going to change
– Stepped care approach?
– Sense-maker of the data deluge?
• “Evidence-based” definition is changing
– New methods for knowing what “works”
– New focus on algorithms
19. Outline
Digital Technologies – What has changed?
What does this mean for evidence-based
practice?
Where are these technologies going?
20. 500,000th App
Accepted on
App Store
2005
2006
Conceive
of a study
Gather
Pilot Data
2007
Submit
Grant
2008
2009
2010
2011
2012
Conduct the study
Receive
Funding
Submit publications
for review
Flickr – Metrix X
21. MILES Study
• Develop theoretically meaningful
smartphone apps for midlife & older
adults
•
Physical activity &
behavior
sedentary
• Passively assess PA & SB
• Feedback for behavior change
Abby King
22. Components
study arms
mConnec
mTrack mSmiles
t
Calorific
Push component
X
X
X
X
Pull component
X
X
X
X
"Glance-able" display
X
X
X
X
Passive activity assessment
X
X
X
X
Real-time feedback
X
X
X
X
Self-monitoring
X
X
X
X
“Help” tab
X
X
X
X
Goal-setting
X
X
Feedback about goals
X
X
Problem-solving
X
X
Reinforcement
X
X
X
Variable reinforcement schedule
X
X
Attachment
X
"Play"
X
"Jack pot" random
reinforcement
X
Hekler et al. 2011, Personal Informatics Workshop at CHI – Design paper
Social norm comparison
X
King, Hekler, et al. 2013 PLoS One, King, Hekler, et al. Manuscript in Preparation
23. min/week of activity at study completion
Physical Activity - 8wk Results
300
250
200
150
Brisk
walking
(min/week)
100
50
MVPA
(min/week)
0
Analytic
Social
Smartphone Apps
Affect
Paired t [60] = 5.3, p
<0.0001
King, Hekler, et al. 2013 PLoS One,
24. ∆ Food Consumption
15
†
Servings pre day
10
5
**
**
†
**
Physical
Activity Apps
Control
0
-5
Processed Sweets Fatty MeatsFatty
Foods
Dairy
VegetablesFruits
-10
Hekler, et al., Manuscript in Preparation.
Diet-tracking
Intervention
App
25. Lessons learned…
• Activity monitoring by phone only is
difficult
• Each intervention had merit, not potent
• “Right” intervention at “right” time and
place?
• RCT experimental design not enough…
26.
27. Nine Questions for Intervention Development
DEFINE THE PROBLEM
1) What is the problem you are targeting?
2) What influences the problem?
3) How can we help tackle the problem?
DESIGN THE INTERVENTION
4) How can technology support?
5) How should individual features work?
6) How is the intervention used and experienced?
DETERMINE IF IT WORKS
7) How are components working to impact the problem?
8) Is the intervention overall producing the desired
outcome?
9) How can the intervention be made self-sustainable?
Klasnja, Hekler, Froehlich, & Buman, Manuscript Submitted for Publication
38. Secondary Analysis – System Identification
Hekler, Buman, Rivera, et al, 2013, Health Education & Behavior.
39. Microrandomization studies
14000
12000
Steps per Day
8000
Week Average
Intervention 4
6000
Intervention 3
Intervention 2
Intervention 1
4000
Measurement
2000
0
1
10
19
28
37
46
55
64
73
82
91
100
109
118
127
136
145
154
163
172
181
190
199
208
217
226
235
244
Steps per day
10000
Days
40. Just in Time Adaptive
mHealth Intervention
Daniel Rivera
Co-PI: Daniel Rivera, ASU
Other Collaborators: Matthew Buman, Marc Adams, & Pedrag
41. Other Adaptive Interventions
SMS Adaptive Intervention
Marc Adams
Reinforcement Learning (CS version) JITAI
Susan Murphy
Image and concept from U of Michigan (Murphy, Tewarli et al and
Pedja Klasnja
Ambuj Tewar
presented athttps://community.isr.umich.edu/public/jitai/Workshop.aspx
43. Digital Tech is
becoming increasingly
powerful.
It will change your
practice… be ready.
Flickr-RansomTech
44. Open Questions…
• What will the role of a Psychologist be
in the future?
– Stepped care approach?
– Sense-maker of the data deluge?
• What will interventions look like in the
future?
• What can we do to ensure this
technology does not become evil?
45. Thank you!
Flickr – veo_
For these slides visit:
www.designinghealth.org
ehekler@asu.edu
@ehekler
Notas do Editor
Discuss the lack of understanding from behavioral scientists on how to really deal with big data and opportunities for setting up “in the wild” studies that could later be harnessed for A/B testing. Nice melding of behavioral science knowledge of randomized controlled trials and HCI’s knowledge on the systems to automate those types of systems in the real-world.
Did a bit of ‘spreading’
Did a bit of ‘spreading’
Did a bit of ‘spreading’
To demonstrate this, let me first describe two models for developing interventions. On the one side, there is the “top-down” approach where content is generated by experts and experts evaluate the information. This is the model that I have been working under and I believe most scientists think about when developing interventions. The strength to this method is that techniques are rigorous but it is also slow. In contrast, another model of intervention development could be called “crowd-sourcing”. In a crowd-sourced solution content is generated by the crowd and evaluated by the crowd. This of course creates a very fast system but it is difficult to know if the information is accurate or the intervention useful. This model, however, is increasingly gaining favor among technologists and thus it is something we scientists must be mindful of. For example, who uses Encyclopedia Britannica’s website? OK, who uses Wikipedia? Wikipedia is powerful because it is crowd-sourced and we must be mindful of that if we want to find way to get technology to use behavioral science.
NOTE, this current draft is just to get a sense of timing and flow on key points to discuss. Formatting on almost all slides will not remain (e.g., likely will NOT have the titles at the top like that).
Science has been a very thoughtful and deliberative process.We move slowly to be “certain” we know something.We are moving so slowly, however, that we are making ourselves obsolete.Take for example the pace of science. Here is a typical timeline for a large NIH-funded grant (the gold-standard for health researchers).Compare this to the pace of technology companies moving.We need to do better and currently, our old ways of thinking about behavior change, including our old theories are frankly, not up to snuff to the challenges and opportunities that mHealth technologies allow us.
In particular, when I was at Stanford, I pitched an idea of developing theoretically meaniungful smartphone apps to increase physical activity and decrease sedentary behavior via the passive tracking of physical activity and sedentary behavior using the phone’s accelerometer to provide feedback to support behavior change. My mentor, Abby King, liked the idea so I organized a team and we ended up submitting and receiving an ARRA Stimulus grant to build this out.
Beyond the common elements, there are also unique elements for the three active applications, as identified here. The key idea study was to parse apart different ways to frame the information about physical activity and sedentary behavior. Rather than labor through this chart, I’m going to show you images from each of the applications to help you get a sense of how they are similar and different.
We have been exploring this in the MILES project, which stands for Mobile interventions for lifestyle exercise at Stanford. As stated earlier, the study is an NIH-funded challenge grant. We are currently finalizing development of our three applications. We plan to start our pilot study in January 2011.
We have been exploring this in the MILES project, which stands for Mobile interventions for lifestyle exercise at Stanford. As stated earlier, the study is an NIH-funded challenge grant. We are currently finalizing development of our three applications. We plan to start our pilot study in January 2011.
We have been exploring this in the MILES project, which stands for Mobile interventions for lifestyle exercise at Stanford. As stated earlier, the study is an NIH-funded challenge grant. We are currently finalizing development of our three applications. We plan to start our pilot study in January 2011.
Following this discussion on this, we started to think more deeply about the entire development model and, after several iterations, we landed on nine core questions, that can be chunked into three broad domains of defining the problem, designing the technology, and then determining if it works. While this might seem obvious, it is interesting that from a discplinary silo approach, many of these points are often not really properly considered. To further flesh this out. We’ve identified these questions for defiing the problem. As part of this, we’ve also articulated in our paper that is currently under review, a variety of methods from a variety discipline to help answer these questions. For designing the technology, we believe these are the important questions, and then finally, for determining if it works, we are interested in these questions.
First, here are the three “glance-able” displays for the applications. Although the information gathered is identical, minutes engaged in sedentary behavior and MVPA, the way we are displaying it is quite different in each app. For the cognitive app, we wanted to frame the information relative to goals as this model assumes that behavior change occurs through active goal-setting and problem-solving through an active “cognitive” process. For the “affect” app, we utilizing a bird “avatar” as the method of tracking your activity. In this app, as you are more active, the bird flies faster, is happier, and becomes more playful. The idea here is that we believe a person would map the bird’s mood, particularly as it feels happier to their own mood and thus create a link up between being more active and feeling better. Finally, for the social app, you will notice that there are multiple stick figures on the home screen. With this design, the idea here is that a person will be motivated to be more active based on the level of activity of other participants in the study via social norm motivations. These glance-able displays set up the differences between the three apps but now I’m going to show you some more specific elements in each.