1. NITA-U Budget Framework PaperPresentation at the
Workshop on Artificial
Intelligence
by
Dr. Fredrick E. Kitoogo (NITA-U)
4th July 2018
Improved Government Performance
through Artificial Intelligence (AI)
2. 2
1. Introduction
2. AI Complementing Govt
Work
3. Use Cases
4. Choices for Application of
AI
5. Strategic Realization & Way
Forward
6. Questions
Presentation Outline
3.
4. Government is the authority that sets rules for a society, helps
its members relate to one another and to others, and keeps it
running smoothly, securely, and peacefully.
Background
1. Human Capital
2. Information
3. Infrastructure
4. Financial
Resources
5. Planning
6. Performance
Measurement
5. A quick analysis of what government (MDA & LG) work day shows that the
most time-consuming task is documenting and recording information
(10% of the total time)
Government Work
Source: Deloitte analysis of Government Work Analysis
MDA
LG
7. A mechanism that introduces innovations in terms of
modernization, transformation, and professionalism for the
improvement of Public institutions and processes.
Public Sector Reform
Emphasis over Time
ImprovementinPerformance
Public Sector Reform in Uganda
Raise
awareness
for need for
change
Identify
Opportunities
Obtain
MDA/LG Buy-
in
Develop PSR
Program
Implement
Changes to
Achieve
Results
8.
9. AI is the simulation of human intelligence processes by machines,
especially Computer Systems. Includes: learning (acquisition of information
and rules for using the information), reasoning (using the rules to reach
approximate or definite conclusions) and self-correction.
What is Artificial Intelligence
10. Rule-
Based/
Expert
Systems
Speech
Recognition
Computer
Vision
Machine
Translation
Machine
Learning
Robotics
Natural
Language
Processing
Complex and difficult task of organizing and
understanding language in a human way (scan
websites for topical discussions, extract terms
and conditions from contracts, identification of
people and places, etc.)
Capture and use expert’s knowledge to
provide answers to tricky but routine
problems. (Judicial and medical decision
support, etc.)
Translates text or speech from
one language to another
(defence, intelligence,
international relations, domestic
applications, etc.)
Domains
Programming with data, computers
learn how to learn, mining info to
discover patterns in data to help
predict the future (traffic flow, crime
pattern, predictions and course of
action, etc.)
Transcribes human speech
automatically and accurately
(dictation, phone assistance,
etc.)
Ability to identify objects, scenes,
and activities in naturally
occurring images (medical and
agric images – disease, crime
scenes, license plates, etc.)
Creation and use of machines
to perform automated physical
functions (drones, robots –
disaster response, airport
handling, health care, etc.)
Practically, there is combined usage of the AI/Cognitive technologies to augment
human intelligence capabilities
11. The AI domains can be organized in three broad categories: Robotics
& Cognitive Automation, Cognitive Insights and Cognitive
engagement.
AI Government Improvement Categories
Problems Mitigated
Client
• Tedious appln
process
• Long wait times
• Language
barriers
Caseworker
• Boring repetitive
work
• Heavy
administrative
burden
• Prioritization
challenges
• Inconsistencies
• Speed
17. Enter Data Manually
Wrong Decisions
Collect Data Manually
CURRENT STATE: Analysing data for Decision Making.
1 2
6
Identify Data Sources
Manually
3Insights Creation
7
5
Ad-hoc Data Entry &
Aggregation & Analysis
4
17
Retype insights into a Word
Doc
18. Right Insights and Decisions
1 Continuous Sensing
and Data Collection
3
4
18
IDEAL STATE: Analysing Data for Decision Making.
2 Manual Data Entry is eliminated
Automatic Data Aggregation, Correlation,
Analysis, Predictions, etc.
19. Current STATE: Citizen Seeking Employment
Applicant delivers application
physically, by mail or email
Hiring Agency contracted
Citizen searches internet, news papers,
offices, and calls colleagues
Background checks conducted
1
6
3
4
7
Recruiting Organization manually
evaluates applications
No verification from NIRA
Bias
The applicant writes several
hundreds applications.2
5 Invitation of eligible applicants for
Interviews
19
Interviews Conducted and
assessed
8 9 Successful and unsuccessful
candidates notified
Complaints and petitions!!!!!
20. Ideal STATE: Citizen Seeking Employment
Applicant delivers application
physically, by mail or email
Hiring Agency contracted
Citizen logs onto a Website and
encounters an Interactive Virtual Assistant
Background checks conducted
1
6
3
4
7
Recruiting Organization
automatically shortlists &
Background Checks
No Bias
The applicant writes several
hundreds applications.2
5 Invitation of eligible applicants for
Interviews
20
Interviews Conducted and
assessed
8 9 Successful and unsuccessful
candidates notified
Happy applicants!!!!!
Jobs, CV, eligibility, Training, Salary,
Training, etc.
21.
22. Three V’s Framework
AI isn't the solution to all Government Challenges, each MDA/LG
Should evaluate the Business Case for each technology.
Value Condition Application Examples
Viable Low to moderate skill: some human
perception required to complete all or
part of a task
Forms processing, first-tier customer service, mail sorting
archives management
Tasks dealing with large datasets Investment advice, medical diagnosis, fraud monitoring
using ML
Rule-based tasks or expertise Scheduling, maintenance operations, complying with
Government Regulations
Valuable Highly skilled workers can focus on
higher value activities
Writing performance & budget reports, e-discovery,
driving/piloting, tabulating data
High labour costs Eligibility determination, answering customer queries,
security/threat detection
Scarce expertise: improved
performance requires AI
Medical diagnosis, aerial surveillance, crime prediction
Vital Industry-standard performance
requires AI
Online drivers license or passport renewal, cyber defence,
criminal investigations, weather predictions
Human labour insufficient to scale
task/service
Fraud detection, patent issuance & IPR protection, disaster
response, text mining
Large backlog: task requires AI Analysis of historical reports, autonomous vehicles and
drones, civic data, etc.
23. Business Impact Choices
AI implementation calls for the need to examine carefully the
business implications of various choices.
Automation Choice Cost Strategy Value Strategy
Replace: repetitive tasks
e.g. mail sorting,
Eliminate Worker Re-assign Worker
Relieve Eliminate routine tasks,
increase productivity,
reduce staff
Redeploy people to high-
value tasks, create more
value for customers
Split-up Breaking up jobs into
pieces and automate some
pieces
Workers are assigned to
specific tasks
Augment and extend Increase performance Increase workers
performance and use to
enhance their skills
24. Notwithstanding the significant potential benefits from
increased data sharing, there may be important risks and trade-offs
that any Government policy should consider to avoid harmful or
unintended consequences..
Potential Risks and Trade-offs
Work Morale
Ensure
mitigation of
embedded
Bias in some
AI tools
Flexibility in
Work Rules
and
collective
bargaining
25.
26. Government AI Strategy - What needs to
be done?
Not every task should be augmented by AI
Government should setup and AI Taskforce to identify common tasks that can be
augmented by AI. Specific teams in MDAs should analyse further specific tasks
that can be augmented by AI. Government could consider developing a
deliberate Government AI Guideline and Strategy to facilitate planned and
controlled implementation
27. Government AI Strategy - What needs to
be done?
MDA expertise in artificial intelligence could boost AI’s Potential.
MDAs shall need a robust AI Workforce to manage the growth and potential of
AI. These experts could serve as the repository of AI knowledge for Government
programs and could work directly of projects when specific expertize is lacking in
any MDA
Government could work with outside experts, especially Universities and
Innovation Hubs.
Universities and Innovation Hubs enormous experience, ongoing research,
development programs and projects. Government can capitalize on those skills
and work together with them to complement knowledge and skills