AI Notes
Learn applications. Medical diagnosis, planning. Expert systems 2024.
Introduction
Expert systems are AI programs that capture the knowledge of human specialists and apply it to solve complex problems within specific domains. Unlike general-purpose AI, expert systems excel in narrow but deep domains where human experts typically take years to develop proficiency. They have been deployed across medicine, engineering, finance, and manufacturing, often matching or exceeding the performance of human experts in their specific areas of competence.
Medical Diagnosis Systems
MYCIN: The Pioneer
MYCIN, developed at Stanford in the 1970s, diagnosed bacterial infections and recommended antibiotics. It demonstrated that rule-based reasoning could achieve expert-level performance.
| IF | (1) The stain of the organism is gram-positive, AND |
| THEN | There is suggestive evidence (0.7) that the organism is Streptococcus |
| Example | Two rules suggest Streptococcus with CF 0.7 and 0.6 |
| - Human experts | 42.5-62.5% |
Modern Medical Expert Systems
| - Knowledge base | 2,400+ diseases, 5,000+ symptoms |
| - Input | Patient symptoms, lab results, demographics |
| - Output | Ranked differential diagnosis list |
| - Updates | Continuously revised with new medical knowledge |
| Rules | IF patient takes Warfarin AND prescribed Aspirin |
| THEN alert | "Major interaction - increased bleeding risk" |
| SUGGEST | "Consider alternative or adjust dosage" |
| SEVERITY | High |
| EVIDENCE | Level A (multiple clinical trials) |
Financial Expert Systems
Engineering and Manufacturing
XCON (R1) System - DEC Computer Configuration
- Configured VAX computer systems (1980s)
- 2,500+ rules about component compatibility
- Saved DEC $40M/year in reduced errors
- Configured systems that would take human experts hours
Example rules
IF: order includes disk drive model X
AND: no disk controller in configuration
THEN: add compatible disk controller model Y
IF: total power consumption > power_supply_capacity × 0.8
THEN: upgrade to next power supply tier
Predictive Maintenance Expert System
IF vibration_frequency in [150-200 Hz]
AND vibration_amplitude > threshold
AND temperature_trend = rising
THEN diagnosis = "bearing wear stage 2"
AND recommended_action = "schedule replacement within 2 weeks"
AND confidence = 0.82
Agriculture Expert Systems
Crop Disease Diagnosis
Input: Leaf color, spots, wilting pattern, weather history, crop type
IF crop = wheat AND spots = "orange pustules" AND location = "leaf surface"
THEN disease = "wheat rust" (CF = 0.9)
SUGGEST: "Apply fungicide within 48 hours, remove affected plants"
Irrigation Scheduling
IF soil_moisture < wilting_point + 20%
AND forecast_rain_48h = false
AND growth_stage = "flowering"
THEN irrigate immediately
AMOUNT: ET_crop × crop_coefficient × area
Advantages and Limitations
Advantages
✓ Explainable: Can show reasoning chain for every conclusion
✓ Consistent: Same inputs always produce same outputs
✓ Available 24/7: No fatigue, no vacation, no mood variations
✓ Knowledge preservation: Expert knowledge survives retirement
✓ Training tool: Teaches reasoning to novices
✓ Handles complexity: Manages thousands of interacting rules
Limitations
✗ Knowledge acquisition bottleneck: Extracting expert knowledge is hard
✗ Brittleness: Fails on cases outside its rule set (no common sense)
✗ Maintenance cost: Rules must be updated as domain knowledge evolves
✗ No learning: Cannot improve from experience (unlike ML)
✗ Narrow scope: Each system handles only one specific domain
✗ Combinatorial explosion: Too many rules become unmanageable
Modern Hybrid Approaches
| 1. ML for pattern recognition | Expert rules for decision logic |
| Example | CNN detects tumor in X-ray (ML) |
| 2. Expert system generates features | ML makes final prediction |
| Example | Rules extract financial ratios |
| Example | Train neural net, then extract rules that approximate it |
Building an Expert System: Design Process
| 1. Knowledge acquisition | Interview domain experts (weeks to months) |
| 2. Knowledge representation | Choose formalism (rules, frames, ontology) |
| 3. Implementation | Build inference engine + knowledge base |
| 4. Validation | Test with known cases, compare to expert opinions |
| 5. Refinement | Add rules for edge cases, adjust certainty factors |
| 6. Deployment | Integrate with existing workflows |
| 7. Maintenance | Continuous knowledge base updates |
Summary
Expert systems remain valuable wherever explainability, consistency, and domain-specific reasoning are required. From medical diagnosis to financial compliance to industrial configuration, they encode human expertise in a form that can be applied reliably and repeatedly. While modern AI has shifted toward learning-based approaches, the principles of expert systems (knowledge representation, inference, explanation) continue to influence how we build trustworthy AI systems that must justify their decisions to human stakeholders.
Exam Focus
Revise definitions, diagrams, examples, and short-answer points for Expert Systems Applications.
Interview Use
Prepare one clear explanation, one practical example, and one common mistake for this Artificial Intelligence topic.
Search Terms
artificial-intelligence, artificial intelligence, artificial, intelligence, expert, systems, system, applications
Related Artificial Intelligence Topics