Layer 1
IFSSA: Harnessing Machine Learning for Social Services

Harnessing Machine Learning for Social Services


We're interested in co-creating a tool that makes intake in the social services sector a transformational & high impact experience. We want to leverage the capabilities of machine learning to help our clients succeed at life changing goals.

The challenges we presently face are…

  1. 1.Our intake system is primarily about assessing eligibility; it passively reinforces an adversarial relationship.
  2. 2.Data collection is a chore; its aggregation & analysis is manual and tedious effort directed at funders, and not a source of real-time practical insights that helps the client in front of you. Data is seen as a tether, not a guiding light.
  3. 3.Data hygiene & data mobility in the sector is low.
  4. 4.Intake is often a highly emotional time, requiring high personal disclosure. It can be (re)-traumatizing for a client.
  5. 5.Data collection practices don't sufficiently factor in mental health.
  6. 6.Intake is the ideal time to identify referrals and options for clients; but it is generally not the best time to present numerous options to a client, and it's easy to forget to follow-up with referrals at subsequent meetings because their nature is more transactional (eg. "here is your food hamper") than relational (eg. "Were you able to connect with…").
  7. 7.The solutions, referrals and options identified for a client depends on which staff member they see, how rushed that person may be, their awareness of resources, new developments, and a host of complex criteria.
  8. 8.We measure deficiencies — need for food hamper, poverty, etc — not skills, aspirations and success — goals, achievement, & impact.

Systemic Racism & Bias

Intake systems reinforce unconscious and explicit bias. They are difficult to monitor, are prone to shortcuts, and hard to evaluate. Machine Learning offers an opportunity to constructively expose bias as well as a systematic approach that reinforces integrity, equalizes referrals and improves access. We realize Machine Learning isn't inherently unbiassed, but we believe it offers an opportunity to help identify and better respond to bias.

Mental Health

Intake systems can have a significant impact on mental health — both for the highly marginalized and vulnerable clients agencies see, as well as the frontline workers who engage in mentally taxing & draining conversations that inflict vicarious trauma. Machine learning can help us design a better process by comparing the efficacy of different approaches in a systematic and rigorous way. The same tools that let us test the effectiveness of an email campaign, A/B testing, psychological nudges, etc, can also help to make intake a better process.


Caseworkers spend as little as 20% of their day on human interaction. Paperwork and other non-interactive tasks consume up to 50 percent of caseworkers’ time.

Source: Unlock the Power of Data for More Effective Social Programs



IFSSA's Preparations

  1. 1.Identifying meaningful indicators & articulating clearly the impact we want to have through a six-month cohort with Dialogues in Action.
  2. 2.Working with the University of Alberta, MacEwan University, and Roundhouse to examine current processes from a Public Health lens and conduct a literature review.
  3. 3.Working with the Edmonton Social Planning Council, University of Calgary, and NorQuest College to do deep-research on long-term clients and the roots of dependency on our services.
  4. 4.Working with Communities United, an umbrella group of North East Edmonton social service organizations to identify approaches for shared referrals and scalable solutions.

What IFSSA can bring

  1. 1.5+ years of client data
    IFSSA has 5 years of data on clients, their usage patterns, demographics, and services accessed, as well as practical experience on data gathering habits, challenges & routines.
  2. 2.Sector experience & partnerships
    IFSSA has 25+ years of social services experience & strong partnerships across the sector that would support adoption and rollout of a robust referral network. IFSSA has been championing common intake within several umbrella groups.
  3. 3.Support building the market
    We believe the creation of an ML based intake system would lead to marketable products with strong revenue potential that can deliver value to non-profits and insights to policy makers, planners and others.
  4. 4.Charitable status
    IFSSA's charitable status allows us to issue in-kind tax receipts and apply for grant funding.
  5. 5.Positive press, employee engagement, employee attraction/ retention
    We believe this project will have numerous positive aspects for our partners, including good press, stronger employee engagement, fresh leads, and new revenue streams.



The Future

Our new intake system needs to…

  1. 1.Move from interrogation→ conversation.
  2. 2.Prioritize mental health and holistic assessment. We need to recognize the assets clients come with, not just deficiencies and demographics. We need to identify tailored referrals based on the client's specific circumstances. eg. skills, primary language, neighbourhood, number of kids, socialization, etc.
  3. 3.Use a systematic line of questioning to identify the goals that will have the highest impact. The LifeWorks Self-Sufficiency Matrix is something we want to build upon.
  4. 4.Facilitate more disciplined practice of ongoing conversations with clients, including follow-ups on goals, referrals, etc.
    eg. Every visit to the food bank should be an opportunity to discuss progress, work on roadblocks, and move towards accomplishing goals. Goals may be as small as increasing positive socialization, a simple budget, getting employment ready, or finding affordable housing.
  5. 5.Automate referrals to clients, and push new opportunities to clients when appropriate.
    eg. Pre-register a client for when a new language skills class is starting for women who speak Arabic and English at level ≤ 5 with 3+ dependents, and no prior EI claims.
    eg. Automatically inform clients when an application for HeadStart opens
  6. 6.Build solutions based on the proliferation of smartphones — the new system should allow an intake to happen anywhere someone can use their smartphone.
  7. 7.Securely transmit client data between organizations and reduce the need for clients to repeat their story again and again.
  8. 8.Incrementally improve, and adapt to changing circumstances nimbly.




Today

Support Worker (SW): I need to ask you a few questions to see if you qualify for our food bank program. What's your household income?

Client: $28,000 per year

SW: How much do you pay for rent?

Client: $950/m

SW: How many dependents do you have?

Client: I have a 6 year old, a 9 year old, a 15 year old, and a parent in my home. We are all struggling.

SW: Are you getting the AISH or the Child Tax Benefit?




Tomorrow

Support Worker (SW): I'm going to use this tool to help guide our conversation. It will make suggestions as we work through some questions. What's your household income?

Client: $28,000 per year

SW: How much do you pay for rent?

Client: $950/m

SW: The tool says you're spending more than 40% of your income on housing and it's suggested a few resources we could explore together. Do you think we should set a goal around finding you affordable housing?

Client: Affordable housing would be a real help.

SW: How many dependents do you have?

Client: I have a 6 year old, a 9 year old, a 15 year old, and a parent in my home. We are all struggling.

SW: I'm sorry to hear that. Have you heard about our youth mentorship program?

Client: No, I'm not sure what that is.

SW: I can send a follow-up text message to you after we address today's issues with some quick videos you can watch as a family. I can also pre-register your kids both for our programs and some with partner organizations.


Is this really a Machine Learning challenge‽ 

Why Machine Learning and not just a good database with a strong rule based engine? Machine Learning offers us the opportunity to…

  1. 1.Optimize service/referral delivery. Use large data to identify the impact of slight changes in timing, phrasing, etc.
  2. 2.Better anticipate/ predict client needs and design preventative services.
  3. 3.Aggregate client data from multiple sources to provide social workers with a more complete picture.
  4. 4.Help articulate and measure impact in more meaningful ways — including facilitating more longitudinal data.


Goals

We want to make the following shifts to the way we work…

1. Data collection as a chore → Data drives insights
2. Assessing Eligibility → Achieving Goals
3. Transactions → High Impact Referrals
4. Measuring Outputs → Measuring Outcomes
5. Reduce stress & Increase impact

Principles

1. Measure what matters
2. Enhance – don't limit – frontline ingenuity
3. Equip clients with powerful choices

About IFSSA…

People come to IFSSA for support with security, safety and growth. IFSSA serves 4500 clients every month, runs Edmonton's second largest food bank, provides aid in financial crisis, supports victims of gender-based violence, settles refugees, and delivers a range of preventative programming for youth. IFSSA's services are open to all.

IFSSA is an Imagine Canada accredited charity, the winner of the Government of Alberta's Inspiration Award for its work combatting domestic violence (2020) and the Canadian Mental Health Associations Professional Service Award (2020).


Quick IFSSA Stats…


Next Steps

We're interested in co-creating tools that harness machine learning to positively change peoples lives. We're open to a variety of approaches and options. We'd like to discuss what co-creation could like.

Omar Yaqub MBA BSc
[email protected]
Servant of Servants (ED), IFSSA
780 695 7477