Mobile Sensing of Smoking Behavior

Smokers will carry an Android-based smartphone, which they are to use as their own for one month. After using the smartphone for two weeks, they will abstain from smoking for 48 hours. The phone will passively sense and record information from onboard sensors and send that information to a central server. When server based algorithms detect a pattern of signals likely associated with smoking behavior, the smoker will be queried regarding their current state (smoking?, not smoking but likely to in the next 10 minutes?, etc.). Likewise, when smokers are about to smoke but were not queried, they can indicate they are about to smoke. This information will be used to update algorithms using machine learning techniques. As such, in this study investigators will gain knowledge that will increase understanding of antecedents of smoking behavior and improve the accuracy with which smoking risk can be detected.

Inclusion criteria

• generally healthy
• between the ages of 18 and 65
• smoking of at least 10 cig/day of a brand delivering ≥ 0.5 mg nicotine (FTC method) for > 1 year AND an afternoon expired CO concentration >5 ppm (to confirm inhalation) or morning urinary cotinine > 100 ng/mL
• willingness to carry and use a study smartphone as their own for one month

Exclusion criteria

• inability to attend all required sessions
• use of smokeless tobacco products
• current use of nicotine replacement therapy or other smoking cessation treatment