How can we protect senior citizens from financial scams and reduce the burden on family members?

University of Washington, Master’s of Human Computer Interaction and Design, graduation project. Sponsored by Blink UX

Time: March – Aug 2016
Role: User Researcher, Project Manager, and Team Leader 

Executive summary

Financial scams against senior citizens are a serious issue, and one for which few effective solutions exist. In our research, we found that family members are typically most concerned about their senior loved ones being defrauded. However, they are usually busy in their own lives and do not have as much time as they would like to check in with senior members of their families.
As a result, we created Voyce, a cloud-based phone fraud detection system that identifies and monitors scam calls, cautions seniors, and alerts family members to suspect activity on seniors’ phones. Voyce comes with a voice-controlled, senior-friendly artificial intelligence assistant named Joyce.

We envision family members will gain peace of mind as their senior loved ones will be better protected from financial scams through the use of our service.


Overall Process


1. Investigate

Desktop Research | Subject Matter Expert Interviews | Competitor Analysis

We learned from secondary research that financial scams pose a major issue for seniors,
one which is getting more prevalent as our society ages.

In 2015, Federal Trade Commission received over 1.5 million financial fraud-related complaints, 37% came from age group 60 and over(This almost certainly understates the extent of scams against senior citizens because scams from this demographic of victims are greatly under reported.)

The issue of financial scam looms large, particularly as the number of older Americans rapidly expands. According to the latest U.S. census, 19% of the U.S. population are over 60The issue looms larger as our society ages.

Both our survey of existing research and our own primary research revealed that most senior scams occur via phone. As a result, we decided to focus in that area to build a solution.


We also learned what key factors make seniors vulnerable to financial scams by interviewing experts in aging and financial fraud:

  • The fact that seniors have more saving, being too polite, and suffering from cognitive decline, and being lonely make them an easy target.  – Leeta Scott, Financial Fraud Helpline, AARP
  • Anyone over 60 years old could suffer from the decline of the following: judgement skill, language, memory, multitasking, problem-solving, and mental flexibility. However, seniors usually don’t realize that.  – Kristoffer Rhoads, Clinical Neuropsychologist, Department of Neurology, University of Washington
  • Senior citizens who are ill, isolated, divorced, or who live away from family have higher chances of becoming victimized due to loneliness and lack of protection. – Kathi Church, Case Manager, Seattle Aging and Disability Services
  • Seniors in most cases, whether wealthy or not, don’t manage their finances or personal assets effectively. They don’t hire financial managers because they don’t believe that they are worth the cost and trouble. – Mattie Taplin, Senior Services Volunteer Manager, Sound Generations

Unfortunately, existing solutions are sub-optimal. Most existing solutions from organizations such as senior centers focus on increasing awareness, which is not so effective for people who suffer from cognitive decline or dementia. Additionally, these solutions generally require social workers to visit in person, making it difficult to implement at scale.

To mitigate scams coming via phone calls, some family members opted their senior relatives into the National Do Not Call Registry. However, callers are in no way bound by the registry, so the system can not stop fraudulent calls.

Call management services such as Truecaller only tell users the call is identified as spam. Seniors are still able to pick up the call and talk to potential scammers.

2. Explore

Field studies | In-depth interviews

To learn more about the issue in detail, to build empathy, and to explore the opportunities, we conducted field visits and in-depth interviews with five senior citizens who have experienced financial scams, as well as five family members of victims.


As we interviewed more and more seniors and family members, I started to identify different patterns and categories of seniors. Eventually, I created a 2×2 matrix to categorize the seniors. The Y axis is the level of judgement / awareness (how much seniors can recognize and deal with financial fraud by themselves); while the X axis represents the level of concern (how concerned family members are about seniors falling victims to financial fraud.)


We decided to focus on seniors that full in Type A and Type D for following reasons:

Type A

Seniors in this group are independent and can manage their own finances with some support from their banks. They are more skeptical and exercise good judgement toward fraud. They also know what to do when they receive calls from scammers. Generally they are not in urgent need of help. However, sometimes they are too confident that they do not set up any fraud prevention mechanisms. This could make them vulnerable to fraud. Another concern for this group is that their cognitive ability to discern and judge might be affected as they age. As a result, we set them as secondary target.

Type B

We have not yet met seniors from this group: maybe there are not too many people like these. We will keep an eye out for seniors who fall into this category in the future.

Type C

There are generally two kinds of situation in this category: 1) The situation is too complicated for us to solve It involves a lot of family issues, complicated relationships, dating scams, etc. Most of these topics are too sensitive and personal, which victims do not want to share. 2) The seniors actually have no money to get scammed for, or a trustworthy family member has taken complete control of seniors cards, bank accounts and personal information. Therefore, we decided not to focus on this group for now.

Type D

Seniors in this group tend to have cognitive impairment and mild dementia. Despite their financial situations being simple, they are highly vulnerable to fraud due to their low awareness and judgement. They are vulnerable to even basic phone scams. This is a strong pain point for the family members, as they are not able to oversee every interaction that can potentially result in the sharing of personal information, and they have limited time and attention to support the seniors. This results in the fact that seniors in this group tend feel that they are burdening their family members. As the need is obvious and there are no effective solutions out there, we aim to set this group as the main target.


Moving forward, we made persona for Type A and Type D seniors, to draw design implications from their pain points and the experience with financial scams. For example, below is the persona for Type D senior – having low awareness of financial scams and receiving high concerns from family members



personasThe personas we created

To summarize, in the exploration phase, I identified different segments of seniors based on their ability to detect and deal with financial fraud  interdependently, and the support and concern they received from their family members. This helped us define the target senior type, and shift the project focus from seniors to family members.

3. Ideate

To move the project forward to the ideation stage, we synthesized findings from all the research – SME, secondary research, competitor analysis, family member interviews, and senior home visits – and concluded three design implications:

1. Seniors have strong desire for independence in finance and in life

According to our SME interviews, there’s a significant complication in senior financial fraud: family members need to keep the elderly protected from fraudulent phone calls, but seniors want to be independent. Our interviews with seniors and family members also suggested that seniors don’t want to be treated like children and it is hard for family members to find a balance.

“My mom refused to let me take care of her bank account until she fell victim to a vacation scam”
– Shane, 45, family member, 75 year old mom

“I help my mom with her mail. She doesn’t like the lack of control”
– Kelly, 45, family member, 80 year old mom

The solution should empower seniors, give them a sense of independence, reduce their dependency on family members.

2. Family members worry more about financial safety than seniors who suffer from cognitive decline do, However, they don’t want to be overwhelmed with details, they want just enough information.

“I am 11 out of 10 concerned about my mom. However, it isn’t an immediate concern of hers because she doesn’t think she would fall for it. However, I would like to know as little as possible about her phone calls”
– Kelly, 45, family member, 80 year old mom

“My mom used to be smart but now I am concerned because of the dementia and her being confused a lot. She does not have any precautions in place.”
– Angela, 51, family member, 79 year old mom

This theme implies an urgent need for another population – family members. We need a solution that will fill in the gap when family members cannot take care of seniors, as well as reduce the burden of seniors and enable senior-family member communication. Family members would like to know that their senior family member is safe and healthy, and would like to know if there’s any emergency, but not be burdened with additional details like call logs, surveillance video, etc.

3. Despite having difficulty using current technologies, most seniors are open to certain kinds of technologies that are easy to use, such as voice control.

The secondary research indicates that many seniors remain largely unattached from online and mobile life – 41% do not use the internet at all, 53% do not have broadband access at home, and 23% do not use cell phone (Pew Research Center, 2014). Furthermore, the majority of our interviewees (four out of five) do not use computers.

“I can’t see the screen well. I tried to ask my grandson to install Cortana because it is more convenient for me.”
– Julee, 74, senior

“I am looking into Alexa by Amazon because you can just ask it questions and it’s just voice controlled.”
– Angela, 51, family member, 79 year old mom

Create a platform that respects seniors’ routines and preferences. The solution should not require seniors to learn a new technology or change their current habits. An example for such platform is a voice-controlled interface. The types of scams and fraud evolve over time, so our solution should be scalable and also be able to evolve in order to be able to identify and prevent future scams.


With the design implications in mind, we then set our design prompt:

“How can we protect seniors from financial scams and reduce the burden on family members?”

We came up with more than sixty initial ideas. We used affinity diagrams to organize the ideas, and prioritized them based on factors such as feasibility and desirability.

Eventually, we decided to move forward with the idea of an AI assistant that detects suspicious calls, cautions seniors through voice-based interactions, and notifies family members when seniors are having conversations with potential scammers.

SystemMapIdea concept map

flowchartThinking about how the system should work

4. Prototype & Evaluate

We turned the selected idea into audio prototype using a text-to-speech software played through a wireless speaker. We also created high-fidelity wireframes for the app that will be used by family members.

prototypeWe played the audio prototypes through a Bluetooth speaker. Click here for an example of the audio prototype

We then evaluated the prototypes with seniors and family members. Main research questions include:

  1. How do seniors and caregivers react to the service?  Can the service solve their issues?
  2. How do seniors want to interact with the AI assistant?  Is the current interaction understandable and easy to use? What are the challenges?
  3. How do caregivers want to interact with the app? Is the current set-up process understandable and easy to use? What are the challenges?
  4. How much information about seniors’ phone activities do caregivers want to know? How much information are seniors comfortable sharing?


Evaluation with seniors

Scenario-based 1:1 interviews and observations. Tested with audio

Three seniors (80, 75, and 77 years old).

We went through three main scenarios – blocking calls directly, listening to voicemail, and adding a trusted ID. By observing how seniors interacted with the audio prototypes, we learned how to improve the voice-based interaction system. We showed the system’s script and asked senior participants to revise it after interacting with the audio prototype. It helped us make the scripts more senior-friendly.




Evaluation with family members


Scenario based 1:1 interviews and observations. Card sorting. Tested with audio and app prototypes.

Two family members (with senior family members that are 66, 86, and 90 years old)

We walked participants through how the system works and played audio clips to show how the AI assistant would interact with scammers, callers and seniors. This helped us learn whether the service is desirable and useful for our potential customer – family members. Participants went through “setup” and “receive notification” scenarios using interactive prototypes on invision. This helped us identify usability issues and see if the flow makes sense to users. At the end of the interview, we asked family members to sort features shown on cards / post-its based on their preferences. The helped us prioritize features.


Key findings – seniors’ interactions with the audio prototype

Due to the lack of experience and understanding of AI voice interactions, it might take some time for seniors to learn how to activate the voice assistant (by saying “Hi_(name of the service)_”). They might start to talk to the system directly. In order to get familiar with the interaction with the AI, seniors need some scaffolding.

In addition, seniors interact with the AI as if they were talking to a real person. They might say “ok / yes” while the AI is still talking, and “thank you” after the AI completes an action.

  • Instead of saying “Hi AI”, “Add Nancy as a trusted ID”, senior participant 1 said “Nancy” directly to the system.
  • AI: “Nancy has been added as a trusted ID”. Senior participant 2: “Thank you”.
  • Senior participant 1: “I wants the conversation with the virtual assitant to be like talking to a real person”

We need to refine the voice interaction model based on seniors’ conversational habits, so that seniors can have effective, intuitive and easy to reassuring can interactions with our system.

Key findings – family members’ interaction with the app prototype

Instead of being able to customize what to monitor and when to be notified, family members do not want to make too many decisions. They want the system to decide for them, with the flexibility to change if they want.

Meanwhile, the level of seniors’ cognitive impairment and the level of security caregivers want to have do not necessarily align.

  • Family member participant 1: “I don’t want to choose these (ScamScore)…can’t it decide for me?”
  • Family member participant 2: “my mom does not suffer from cognitive decline, and she has a good sense of judgement, but she still answers calls which she knows are telemarketers. Therefore I still want the system to monitor everything and notify me.”

Reduce the decisions family members need to make during the setup process. Provide default settings with the flexibility to customize later on. In the setup process, instead of asking for the senior’s cognitive decline level, ask the level of security users want.

Usability Issues

Audio prototype

  • The AI can play beeping sounds when seniors receive voicemails. This is what landlines currently do and seniors are familiar with it.
  • When adding Trusted ID, consider the situation when the names are long and uncommon, and where users may need to spell them out.

App prototype

  • Caregivers were a bit confused by the “voicemail” feature. We will need to add additional on-boarding slides.
  • Certain terms are app-specific and don’t need to be. For example, “Trusted ID” can just be “Contacts”.
  • Certain terminology used may not be well understood. For example, “machine learning”.

 5. Production & Final product specification


Voyce, the finalized product, is a voiced based AI system that help family members protect their senior loved ones from phone scams.

There are four components of Voyce: a cloud-based scam database, that identifies calls and fraudulent phone conversations, a senior-friendly voice interface that cautions senior citizens, and an app that alerts family members.




1. Identify and analyze incoming calls

When a number calls in, Voyce checks if the caller is in the scammer database, the senior’s contact list, and if it is a credible organization. Then Voyce connects the caller to the senior.

For callers that are not in the senior’s contact list or from credible organizations, Voyce categorizes them as “unidentified” and directs the caller to leave a voice message for the senior. Voyce then analyzes the voice messages.

2. Monitor fraudulent conversations (based on the level of security selected)

Voyce also monitors every caller’s phone conversations with the senior, checking if the conversation contains fraudulent content. When combined with other suspicious activities of the caller (e.g., spoofed caller ID), Voyce can spot scammers who are not in the scammer database.

Design principle:
Minimize user input. Most of the activities happen in the background and require no awareness from users.


We designed a voice-based interaction model, with an AI assistant named “Joyce” for seniors to communicate with Voyce. Seniors could talk to Joyce through either their own phone devices (landlines, flip phones and smartphones), or the physical device provided by Voyce.

When Joyce detects a potential scammer, it will caution seniors to:

  • Avoid revealing personal information
  • Check in with family and friends
  • Think logically and calmly before taking actions

It will also calm seniors down by letting them know it is unlikely to receive free offers or be contacted by organizations such as IRS.

Design principle:
The interaction model should be simple, intuitive, and senior-friendly.


Voyce also comes with a family member-facing app. The app guides them through the setup process and includes features such as personalized settings and notifications when seniors are having conversations with potential scammers or receiving suspicious voicemails.

Design principle:
Provide critical but manageable information to keep users posted and help them make decisions.

app2An excerpt of the set up screens

appFamily members receive notifications with each activity, including scam score, suspicious words, recorded conversations, and suggested actions.


Voyce lives in the cloud. It collects data from Truecaller, Nomorobo, and Federal Trade Commission to identify scammers. If callers call on behalf of an organization, Voyce retrieves data from to analyze the credibility of the organization. Voyce also has its own database to monitor and analyze phone conversations. Voyce learns from its users’ reactions to certain caller IDs and updates its database accordingly.

Scam Score

Voyce gives each phone number a “scam score” that increases as callers are potentially more fraudulent. An overall scam score is generated by weighing five key factors:

  1. Content of each phone conversation
  2. Calling frequency (to Voyce members)
  3. The caller’s calling location
  4. Amount of users who report the caller
  5. For callers who are affiliated with an organization, the credibility of organizations mentioned by the caller.

Spotting fraudulent phone conversations

Fraudulent phone conversations tend to contain high amount of trigger words. For example, if a scammer is offering an incentive, he is more likely to use terms such as “guarantee” and “free.” If the scammer is framing his offer in a time-limited fashion, he is more likely to use terms that deliver a sense of urgency. The scammer might also mention actions such as “wire” or “transfer.” With machine learning techniques, we could model normal and fraudulent phone conversations to detect an anomaly in seniors’ phone conversations.

TriggerWordsWe organized a list of fraudulent terms used in a phone conversation.

Next step

Currently, Voyce is designed to collaborate with organizations including FTC (Federal Trade Commission), AARP (American Association of Retired Persons) and Truecaller. Moving forward, we plan to collaborate with organizations such as banks, telecommunication companies, Internet phone services, the IRS, Medicare, and senior centers, to provide access to our product and better cater to their specific needs.


TeamPhoto with the team

SponsorPhoto with our sponsor, CEO of Blink UX

Check the process book (pdf) for more detailed process and product specification.

Other Projects