Date | November 2021 | Marks available | 2 | Reference code | 21N.1.SL.TZ0.4 |
Level | SL | Paper | 1 | Time zone | no time zone |
Command term | Outline | Question number | 4 | Adapted from | N/A |
Question
Artificial intelligence (AI) predicts earthquakes
Scientists at universities are developing a machine learning system to detect the warning signs that an earthquake is likely to occur. Using pattern-recognition algorithms similar to those in image and speech recognition, the system would be able to predict earthquakes a few days before they occur. Using machine learning, researchers will be able to run earthquake analysis 500 times faster than they could previously.
It is also possible to use machine learning to predict where the aftershocks* of an earthquake may occur (see Figure 4).
Figure 4: Prediction of aftershocks using machine learning
* aftershock: a smaller earthquake that follows a large earthquake
Identify two characteristics of pattern recognition.
Outline one reason why it might be difficult to collect accurate data that can be used for predicting earthquakes.
Identify two characteristics of an algorithm.
Scientists used visualizations, such as in Figure 4, to present the information about predicted earthquake activity.
Analyse the decision to use visualizations.
Discuss the advantages and disadvantages of using machine learning to predict natural disasters like earthquakes.
Markscheme
Answers may include:
- Identifies familiar patterns and objects.
- Recognize shapes and objects from different angles.
- Uses mathematical methods.
- Is a branch of machine learning.
Award [1] for each characteristic of machine learning identified up to [2].
Answers may include:
- Data might be limited / not enough data points collected – may reduce the effectiveness of the machine-learning model.
- Reliability of hardware – e.g., sensors/software used in collecting the data.
Award [1] for a limitation of collecting data for predicting earthquakes and [1] for a development of that limitation up to [2].
Answers may include:
- A set of rules…
- that are followed by a computer in problem-solving.
- A sequence of unambiguous instructions…
- with a finite number of steps…
- that are clear and essential.
Award [1] for each characteristic of an algorithm identified up to [2].
Answers may include:
Advantages of visualizations:
- Presents complex information clearly.
- Makes it easier for non-experts to understand patterns/trends/anomalies in the data.
- Visualizations are easier to process than text.
- May lead to more efficient / more rapid decision-making.
Disadvantages of visualizations:
- May present an overly superficial picture that may not address the key patterns/trends/anomalies in the data.
- The design of the visualization may introduce data bias (either intentional or unintentional) that gives misleading information.
- The visualization may not be effective.
- May not give sufficient information on its own to make an appropriate decision
Answers may include:
Reasons why machine learning should be used to predict natural disasters:
- Machine learning learns from previous experiences (systems).
- It can easily analyse patterns from previous experiences (systems).
- AI can be trained to analyse patterns of natural disaster to predict future occurrences.
- Access to vast amounts of data can make predictions more accurate (algorithms, systems).
- Access to worldwide data will lead to better predictions (values).
- The system becomes more accurate as more data is collected over time (systems).
- It can analyse large amounts of data, so decisions will be based on greater information (systems).
- It can detect non-linear relationships within data and provide more information.
- It can detect all possible interactions between variables.
- It can eliminate human error – incorrect information.
- It can react faster than a human to incoming data.
- The recommendations are likely to be based on the greatest data set possible.
- They will be updated instantaneously.
Reasons why machine learning should not be used to predict natural disasters:
- The quantity of data collected may be limited – not enough data to make predictions.
- The algorithm may not be accurate – causes errors in predictions.
- Predictions may not be accurate / reliable due to geographical differences/locations.
- Prediction cannot take into account new scenarios (for the first time) / missing data.
- The recommendations may be generic / be based on algorithmic biases and not be sufficiently customized to each natural disaster (values).
- There may be particular characteristics of the natural disaster that the AI system may not be able to understand (systems).
In part (c) of this question it is expected there will be a balance between the terminology related to digital systems and the terminology related to social and ethical impacts.
Keywords: natural disasters, environment, data, algorithms, accuracy, reliability, change, systems, values, ethics
Refer to SL/HL paper 1, part c markbands when awarding marks. These can be found under the "Your tests" tab > supplemental materials > Digital society markbands and guidance document.