Over the past decade, the investigators have developed and tested a toll with the potential to enhance PROs at the individual level- embodied conversational agents (ECA), which are computer characters that simulate face-to-face conversation using voice, hand gestures, gaze cues, and other nonverbal behavior. The investigators have successfully used ECAs in behavioral interventions for populations with limited health literacy, elderly patients, and patients with cancer. Face-to-face encounters, in conjunction with written instructions supported by pictures, remains one of the best methods for communicating information in general but is particularly effective for individuals with limited health literacy. The investigators have also demonstrated that ECAs can be used as valid alternatives to standard paper-based surveys for substance use screening, and that the display of empathy for patients and other relational behavior by ECAs leads to increased engagement by patients over time.

There is increasing recognition of the important of patient-reported outcomes (PROs) for assessment of health-related quality of life in both research and clinical settings. PROs are especially important for inherently subjective but crucial clinical phenomena such as pain, mood, and fatigue. PROs can also provide valid assessments of health where performance-based objective measures are possible, but cost prohibitive or complicated. Despite much research on PROs, concerns about reliability and validity persist, especially when used at the individual level, particularly among patients who may struggle to understand PRO questions, such as those with low health literacy.

Over the past decade, the investigators have developed and tested a tool with the potential to enhance PROs at the individual level - embodied conversational agents (ECA), which are computer characters that simulate face-to-face conversation using voice, hand gestures, gaze cues and other nonverbal behavior. The investigators have successfully used ECAs in behavioral interventions for populations with limited health literacy, elderly patients, and patients with cancer. Face-to-face encounters, in conjunction with written instructions supported by pictures, remains one of the best methods for communicating information in general, but is particularly effective for individuals with limited health literacy. The investigators have also demonstrated that ECAs can be used as valid alternatives to standard paper-based surveys for substance use screening, and that the display of empathy for patients and other relational behavior by ECAs leads to increased engagement by patients over time.

The goal of this study is to adapt our prior work on ECAs to produce ECA-PRO, a framework for administering PROs over time using an ECA. ECA-PRO will be used to administer PROs from the Patient-Reported Outcomes Measurement Information System (PROMIS), as well as measures from the Common Terminology Criteria for Adverse Events (CTCAE). The plan is to administer PROMIS measures relevant to cancer including physical functioning, fatigue, depression, anxiety, and pain interference. This study will test the system in an RCT with 100 patients receiving chemotherapy and/or radiation therapy for gastrointestinal or head and neck cancer. Participants will be randomized to provide data for 6 weeks using ECA-PRO (n=50) versus standard web administration (REDCap) (n=50).

Inclusion Criteria:

• At least 18 years of age or older
• Able to read and speak english
• Can independently consent
• Must have adequate corrected vision to use the ECA system (based on a 1 minute ECA functional screener)
• Must have adequate hearing to use the ECA system.
• Has a diagnosis of gastrointestinal cancer or head and neck cancer for which they are currently receiving chemotherapy and/or radiation treatment.
• Provider subjects must be practicing oncologists at BMC.

Exclusion Criteria:

• If the subject is incarcerated
• If the subject plans to leave the Boston area in the next 6 weeks
• Is not able to use the ECA screener

The team statistician will be blinded to experimental condition to reduce the possibility of bias. Only after analyses have been completed will the blinds be revealed.