# US 2021/0342704 A1 · System and Method for Detecting Misinformation and Fake News via Network Analysis

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2021/0342704 A1
- Document ID: US20210342704A1
- Status: Published application
- Kind: A1
- Application number: 17/321,395
- Filed: 2021-05-14
- Published: 2021-11-04
- Earliest claimed: 2018-11-14
- Inventors: Elan Pavlov
- Web page: https://patentgrep.com/patent/US20210342704A1

## CPC

- G06N 5/02: Knowledge representation; Symbolic representation
- G06N 5/022: Knowledge engineering; Knowledge acquisition
- G06F 16/313: {Selection or weighting of terms for indexing}
- G06F 16/335: Filtering based on additional data, e.g. user or group profiles (filtering in web context G06F16/9535, G06F16/9536)
- G06F 16/953: Querying, e.g. by the use of web search engines
- G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks
- G06N 20/00: Machine learning
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

A method for detection of misinformation without the need to analyze any articles (HINTS) includes forming a mixed graph containing at least two different node types, such as users and articles, with edges between users and articles, and with user weights for user nodes and article weights for article nodes. Seed nodes are planted for at least one user node and at least one article node. User weights and article weights are manually assigned to the seed nodes, then neighborhoods are defined for the seed nodes. A HITS-like algorithm is then run for a predetermined number of rounds updating both people and articles, while keeping the weights of the seed nodes constant to converge the graph for the weights of articles and users. Finally, a set of highest weights for users and/or articles is outputted and possible remedial action can be taken.

## Claims

### Claim 1

A method for detection of misinformation, comprising:
identifying which of a set of users interact with a specific item of content having undetermined or not fully-determined reliability;
evaluating the reliability of the users that interact with the item of content; and
evaluating the reliability of the item of content based on the reliability of the users in order to determine whether the item of content is misinformation.

### Claim 2 (depends on 1)

The method of claim 1, wherein the reliability of the users is evaluated by assessing the probability that each user will interact with misinformation.

### Claim 3 (depends on 2)

The method of claim 2, wherein the probability that each user will interact with misinformation is calculated based on previous interactions of the user with previously-labeled content having previously-determined reliability.

### Claim 4 (depends on 3)

The method of claim 3, wherein the calculation of probability is performed using a graph.

### Claim 5 (depends on 4)

The method of claim 4, wherein the calculation of probability is performed using the graph by the steps of:
forming a mixed graph containing at least two different node types, called users and articles, with edges between users and articles, with user weights for user nodes and article weights for article nodes, wherein the user nodes represent individual users in the set of users and the article nodes represent individual items of content;
planting at least one seed user node and at least one seed article node into the mixed graph;
manually assigning user weights and article weights to the seed nodes;
defining neighborhoods of the seed nodes;
running a HITS-like algorithm for a predetermined number of rounds, updating both users and articles while keeping the weights of the seed nodes constant to converge the mixed graph for the weights of articles and users; and
outputting a set of highest weights for at least one of users or articles.

### Claim 6 (depends on 5)

The method of claim 5, further comprising updating a first side of the mixed graph faster than a second side of the mixed graph.

### Claim 7 (depends on 6)

The method of claim 6, wherein the first side of the mixed graph is articles and the second side of the mixed graph is users.

### Claim 8 (depends on 5)

The method of claim 5, wherein user weights are only updated when a new user appears.

### Claim 9 (depends on 5)

The method of claim 5, wherein no article is analyzed.

### Claim 10 (depends on 5)

The method of claim 5, wherein user weights are determined by comparison with a control group.

### Claim 11 (depends on 5)

The method of claim 5, wherein article weights are determined by user input.

### Claim 12 (depends on 5)

The method of claim 5, wherein propagation is stopped based on interaction with specific predetermined users.

### Claim 13 (depends on 5)

The method of claim 5, further comprising assigning negative links between users and articles that represent a lack of an expected association between a particular user and a particular article.

### Claim 14 (depends on 5)

The method of claim 5, further comprising normalization and fixed values in the mixed graph taken from manual input.

### Claim 15 (depends on 5)

The method of claim 5, wherein the mixed graph has more than two node types.

### Claim 16 (depends on 5)

The method of claim 5, further comprising using implicit human interactions to propagate content labels.

### Claim 17 (depends on 5)

The method of claim 5, further comprising aggregation of inputs from multiple users to generate content labels.

### Claim 18 (depends on 17)

The method of claim 17, wherein the content labels are aggregated non-linearly.

### Claim 19 (depends on 17)

The method of claim 17, wherein the content labels are of trustworthiness metrics.

### Claim 20 (depends on 17)

The method of claim 17, wherein articles may be updated more often than users.

### Claim 21

A computer program product for detection of misinformation, the computer program product comprising a non-transitory computer-readable medium comprising a set of instructions stored on the non-transitory computer-readable medium and executed on a computing device, the instructions being for:
identifying which of a set of users interact with a specific item of content having undetermined or not fully-determined reliability;
evaluating the reliability of the users that interact with the item of content; and
evaluating the reliability of the item of content based on the reliability of the users in order to determine whether the item of content is misinformation.

### Claim 22 (depends on 21)

The computer program product of claim 21, further comprising instructions for evaluating the reliability of the users by assessing the probability that each user will interact with misinformation.

### Claim 23 (depends on 22)

The computer program product of claim 22, further comprising instructions for calculating the probability that each user will interact with misinformation based on previous interactions of the user with previously-labeled content having previously-determined reliability.

### Claim 24 (depends on 23)

The computer program product of claim 23, further comprising instructions for calculating the probability using a graph.

### Claim 25 (depends on 24)

The computer program product of claim 24, further comprising instructions for:
forming a mixed graph containing at least two different node types, called users and articles, with edges between users and articles, with user weights for user nodes and article weights for article nodes, wherein the user nodes represent individual users in the set of users and the article nodes represent individual items of content;
planting at least one seed user node and at least one seed article node into the mixed graph;
manually assigning user weights and article weights to the seed nodes;
defining neighborhoods of the seed nodes;
running a HITS-like algorithm for a predetermined number of rounds, updating both users and articles while keeping the weights of the seed nodes constant to converge the mixed graph for the weights of articles and users; and
outputting a set of highest weights for at least one of users or articles. wherein, said first, second, third, fourth, fifth and sixth program instructions are stored on said non-transitory computer readable medium and executed on a computing device.
