Chapter 9

Crowdsourcing Intelligence

Crowdsourcing platforms can help you collect actionable intelligence from the most hostile populations and the most remote areas. The process for building and running crowdsourcing platforms that collect intelligence as opposed to accomplishing other objectives is relatively simple. However, the process can vary significantly depending on the target audience and the type of information you want to collect. This chapter goes over the process by explaining what is possible through crowdsourcing and what is not possible, and shows how intelligence platforms differ from platforms used to accomplish other or mixed objectives. It also walks through how to build and run platforms that collect intelligence for analyzing hard-to-reach populations and helping design other crowdsourcing platforms through SMS, helping law enforcement encourage citizens to identify criminal suspects through smartphones, and helping with crisis response through crowdmap websites.

Understanding the Scope of Crowdsourced Intelligence

Crowdsourced intelligence is intelligence derived specifically through custom crowdsourcing platforms, known as intelligence collection platforms, as opposed to from social media in general. It is another form of human intelligence that veers into the territory of signal intelligence. You can apply crowdsourced intelligence to solve a variety of mission sets, ranging from improving disaster relief to identifying and tracking illicit actors and behaviors. Specifically, you can use it to do the types of analyses discussed in Chapter 5 and Chapter 6 of this book, and even the advanced analytics we describe in Chapter 12.

Uniqueness of Crowdsourced Intelligence

Crowdsourced intelligence differs from other types of intelligence in two ways. The first is obviously how you gather it. The second is the form of the data or, simply, what the intelligence looks like. By definition, all crowdsourced intelligence is delivered via different social media technologies. Each technology differs in the type, amount, and format of data it transmits. For instance, Twitter only allows people to share pictures and links, and only 140 characters of text. Thus, the form of the crowdsourced intelligence is heavily dependent on the type of social media technologies you use and the type of platform you build. Table 9.1 lists what form of data you should expect for each type of platform.

Table 9.1 Data Form by Platform Type

Direct and Indirect Crowdsourced Intelligence

Crowdsourced intelligence comes in two types: direct and indirect. Direct intelligence directly and explicitly tells you about a place, an object, or a person. In other words, the intelligence is direct when a person provides you with information about what he or she sees or you receive data about a place, an object, or a person. On the other hand, intelligence is indirect when you have to sift through acquired data that may seem irrelevant to infer things about a place, an object, or a person.

Direct intelligence is more readily useful but harder to get. It is obvious, tells you exactly what you need to know, and requires less analysis. For example, say you are interested in the movements of an extremist violent group in a certain neighborhood. Individuals in the neighborhood use your crowdsourcing platform to tell you where members of the group live, what churches, restaurants, and meeting places they go to, and when they go there. The intelligence you receive is direct. You may have to compile it and verify it from other sources, but you do not need to do much more analysis on it to figure out the movement of the group in the neighborhood. However, getting such clean and useful intelligence is difficult. Generally, the more direct the intelligence, the more you may have to incentivize people to provide it. People will share critical information with someone they do not know, but it takes rewards or time to convince them to do so. Also, keep in mind that in some cases individuals who provide you with direct intelligence are putting themselves at a greater risk. In the example, say that the individual providing you with the information is actually a member of the group. If the group comes across pieces of very sensitive and complete information about their movements, they may correctly reason that someone from within their group is supplying the information. They may then seek out the mole and punish him or her harshly. However, if they come across information that is indirect and not obviously referencing their movements, the group may have a more difficult time finding out who is supplying the information. Due to the difficulty of acquiring direct intelligence, you should expect to receive it only when the incentives are high enough for potential sources to take the risks, or when the target audience approves of your objective.

In contrast, indirect intelligence is less readily useful but easier to get. You have to infer the information you need from lots of relevant and irrelevant intelligence, which takes time and analytical resources. Using the same example, say you are trying to acquire information about the group's movement through indirect intelligence. You can then ask individuals in the neighborhood about how they feel toward the group. Numerous individuals then tell you that they feel scared when they see several members of the group together. You can then ask individuals in different parts of the neighborhood about how they feel at certain times of the day. If their feelings of insecurity spike at certain times of day and that pattern holds true over a long period of time, you can then infer that the group likely appears at that location at that time of the day. You can then compare the locations of individuals who are experiencing insecurity spikes at different times and map the group's movement. Your assumptions and conclusions may not be completely correct, but you at least get some information about the group's movement. Generally, sources are more likely and willing to give up indirect information. The information they are providing is not as sensitive, and in some cases may even appear completely benign and irrelevant to your ultimate objective. Because the risk to the potential sources is low, the incentives they need to overcome the risk will also likely be low. Due to the relative ease of acquiring indirect intelligence but the difficulty of making sense of it, you should expect to use indirect intelligence when the target audience is unsupportive or scared of your ultimate objective, when you do not have much to incentivize them with, and when you have the time and resources to infer conclusions. In practice, you often need to collect both types and use one to help collect and make sense of the other. Figure 9.1 provides a graphical summary of the differences between direct and indirect intelligence.

Figure 9.1 Direct and indirect intelligence

Appreciating the Limits of Crowdsourcing Intelligence

Collecting intelligence through crowdsourcing platforms can significantly bolster your missions, but it is not a panacea. It will not always get you the information you need or do it in a cost-effective and efficient manner. Like other intelligence collection methods, it too can fail, produce costly or dangerous embarrassments, and prove useless. It has a few disadvantages that, if not addressed, can limit its utility and power.

Uncooperative Target Audience

The most obvious disadvantage is that the success of every intelligence collection crowdsourcing platform depends heavily on its target audience. Some audiences will simply not cooperate with you or give you the data you need in a timely manner. Understanding the target audience and structuring the platform and incentive structure in a way that appeals to them will significantly increase the likelihood that they cooperate with you. Unfortunately, no platform type or set of incentives can guarantee cooperation. Sometimes, the risk will be too high for potential participants. For example, it is safe to assume that people in Mexico in 2012, in areas that are currently being ravaged by vicious drug cartel violence, will be very unlikely to provide intelligence about drug cartels in their areas. Drug cartels in such areas regularly and brutally kill people who provide sensitive information about them to authorities. Few people will cooperate and want to put themselves and their families in danger. You may still find a few and incentivize them sufficiently, but the chances of incentivizing a lot of people to cooperate are low.

Higher Risk for Participants

Additionally, in some cases the risk is higher for participants to cooperate through crowdsourcing platforms than through more traditional Human Intelligence methods. Unless your target audience consists of only a few people, word can eventually get out that the platform exists. You can take steps to minimize the possibility that adversaries find out about your platform, but you can never eliminate it. Platforms that crowdsource direct intelligence about violent illicit actors from participants living in unstable regions can especially endanger the lives of the participants. Also, if word gets out about the platform and leads to attacks against the participants, then word will probably get out about the attacks. Increased scrutiny can lead to major fallout such as stories about your failed platform and objective on the front page of the Washington Post or Le Monde. Think about such worst-case scenarios when designing platforms, but do not become paralyzed by them. Risk is not an excuse for inaction. In the case studies, we discuss some ways of helping participants maintain anonymity. In the most dangerous conditions, the best way may be to create a platform that crowdsources only indirect intelligence.

Misinformation

Like intelligence collected via traditional HUMINT methods, crowdsourced intelligence may be false. Adversaries may infiltrate your platform and feed you misinformation. Or, individuals who are not affiliated with adversaries but do not know anything about them either may decide to participate simply because they want the extrinsic incentives. They may submit false information that is usually sensational so they can get your attention and win lucrative rewards. Some participants may tell the truth initially, but start to lie over time. Some become addicted to the extrinsic and intrinsic incentives, such as the need to be useful or stay involved, and start to fabricate their information. However, you can minimize the effect of misinformation by having a large and diverse sample of participants. You can then use information from a variety of sources to validate suspicious information. If someone tells you something that contradicts what 200 other people say, it is probably false. On the other hand, one person may know something that others do not. In such cases of extreme outliers, look for third-party, independent verification. Go into the project with the attitude, “if it's too good to be true, it probably is.” A healthy skepticism will keep you on your toes.

Misinformation not only affects you, but it can also affect other participants and the target audience at large. If your platform allows participants to see submissions from others, an adversarial participant may use the platform to cause confusion and chaos. For instance, in Veracruz, where drug cartel violence is high, the local populace takes to Twitter to inform their communities about violence and possible threats. In the summer of 2011, two Veracruz residents tweeted that drug cartel members were kidnapping children from schools. Parents read the tweet and panicked. Afraid for their children's lives, they rushed out to pull their children from school, generating chaos across the city. The warning, however, was false and the children were safe. However, the panicky chaos among the parents resulted in 26 car accidents.¹

Misinterpretation

Even if you manage to validate all the crowdsourced intelligence, you may still misinterpret it. The chances of misinterpreting data are especially high when you are dealing with indirect intelligence. Often, the cause for misinterpretation will be mistranslations. Use native translators or very sophisticated automated translation tools when translating incoming intelligence. Also, much of crowdsourced intelligence is text-based. As anyone who has sent a sarcastic e-mail knows, sarcasm is often lost in text. Regardless of claims, no existing natural-language processing tools can accurately detect sarcasm, especially in less popular languages. Misinterpretation may also come about if you do not query your participants properly. If the questions you pose are unclear, participants will interpret them differently and respond incorrectly. The best way to minimize misinterpretations is to hire area experts to help you craft your queries and study the responses.

Hawthorne Effect

Lastly, another disadvantage of crowdsourcing intelligence is related to the Hawthorne effect, which is the phenomenon that the behavior of a participant during a study may change because the participant is engaging in the study. In our case, the very act of crowdsourcing intelligence and how you do it may change the nature of the intelligence the participants provide and may even change the participants' environment. For example, if you introduce dumb phones into a community so that you can crowdsource through them, you may fundamentally change how the participants communicate with each other and thus their lives. The ideal way to collect intelligence in most cases is to do so by least disturbing the area in question and the lives of the people involved. However, involvement in a crowdsourcing platform increases the potential for disturbance. In some cases, you may have to provide technology such as cell phones to the target audience so they can participate. Such an introduction of technology can drastically change lifestyles and cultures. Thus, it can affect your target audience analysis and force you to change your platform design.

Also, some intelligence collection platforms offer social networking features that help participants build and change relationships, which can affect the types of people they come across and the sort of information they acquire and share. The effects can be positive or negative for your objective. A participant may make ties with people through the platform who encourage him or her to share information with you. Conversely, the participant may make ties with people who discourage information sharing. Generally, the more technology and social networking features you introduce, the more likely you will affect the lives of the participants and, thus, their ability and willingness to share information with you. In some cases, you may want to minimize the effects, but in some you may want to bolster it. Keep the Hawthorne effect in mind when assessing the target audience before, during, and after the deployment of crowdsourcing platforms.

Tweaking the Process for Intelligence Collection Platforms

Despite the few disadvantages, crowdsourcing platforms still offer a tremendous ability to collect intelligence from seemingly unreachable populations and areas. They are unique platforms useful for accomplishing a specific objective, and thus differ from platforms for other objectives. In general, Chapter 8 describes the process for building and launching them. However, you need to tweak the process and recognize the differences between intelligence collection platforms and crowdsourcing platforms for other objectives.

Usually, intelligence collection platforms are run for shorter periods of time than solution and influence platforms. However, depending on the missions you are supporting, you may need to run intelligence collection platforms for longer periods. Incentivizing people to provide a piece of information does not involve time-consuming activities such as creating social networks and relationships, or even fostering much trust between you and them. As long as you can effectively communicate the fact that people will be safe and get rewarded for the information they provide, people will start providing intelligence. Because intelligence platforms require less time to get going, they require fewer resources to maintain over a longer period of time. You can then save on resources, or use the savings to expand your target audience or make the extrinsic incentives more attractive.

Also, intelligence collection platforms do not require elaborate social networking and messaging features. In some cases, you will not want participants to communicate with each other so you can better protect their identities. Reduced interaction between the participants, and even between the participants and you, reduces the need for moderators. For intelligence platforms that collect indirect intelligence or those that evolve out of influence platforms (as we describe in Chapter 11), you may need to integrate social networking and messaging features.

Finally, we expect that you will use intelligence collection platforms primarily to collect intelligence from rural and poor areas that are hard-to-reach. In these areas, populations are more likely to use dumb phones or access the Internet through cyber cafes. Thus, your platforms will likely be based on SMS networks and websites, as opposed to smartphone applications.

Now that you know where to tweak the process, you can start building and running intelligence collection platforms. In the subsequent sections, we walk through how to build and run several platforms that differ in type, resources budget, target audience, and in the missions they can support. For ease of understanding and comparison, the steps of the walkthroughs follow the structure of the process in Chapter 8.

Collect Intelligence from Hard-to-Reach Areas through SMS

You may need to launch crowdsourcing platforms or other efforts in places where you do not know much about the population. Generally, the more remote the area, the harder it will be to get information about the people there and the more it will cost. Fortunately, SMS technology is becoming ubiquitous, even in the most hard-to-reach and poorest areas. Even if the population there is not already using SMS, you can provide them with dumb phones and they can learn how to use them fairly quickly. Therefore, SMS-based intelligence collection platforms are ideally suited for collecting information about hard-to-reach and relatively unknown populations.

Increasingly, open source technologies such as FrontlineSMS and RapidSMS are making it easier than ever to implement SMS-based intelligence collection platforms, otherwise known as pilot platforms. This walkthrough describes how to use SMS technologies to build and deploy a pilot platform to crowdsource intelligence about target audiences for population analysis or for the design and deployment of more elaborate crowdsourcing platforms.

Define Objective and Scope

Imagine that you are a member of a Western government and you need to collect information about the target audience in Somalia, a place that is not exactly hospitable to Westerners. You need to collect information about the target audience to aid the design of a long-term crowdsourcing platform that collects commodity price information and opinions from residents in a hostile region of Somalia. You then intend to look at correlations between the prices and the opinions to see if you can use the prices as an indicator of instability and insecurity in the region. Initially, you want to focus on areas that are currently stable so you can collect data to establish a baseline correlation (or lack thereof) between prices and opinions. You also want to see if Somalis are willing to text you information about prices. Your budget is fairly small and its enlargement depends on your initial success. For example, you may have only enough money to purchase and distribute cell phones or to travel to your region of interest. So, you intend to focus on a few stable cities at first with the hopes of expanding to less stable areas later. Gaining success will be easier in the more stable areas and, because this is your first crowdsourcing platform, it is easier to start out in a relatively safer area.

Currently, you need to collect two pieces of information about the target audience. The first is an accurate database of phone numbers in the target region that you can send an SMS message to and ask for information.

The second is identifying which of the phone numbers belong to people who are knowledgeable about commodity prices, such as farmers, merchants, and businessmen. With your limited budget in mind, you have the following three options for getting the information:

After reviewing your options, you decide to try the third option first because you really want to assemble a list of phone numbers that you can access at any time. Meanwhile, you seek out a marketing company that has access in your region of interest so you can execute the second option later. The first option is too cost-prohibitive. For now, your objective is to crowdsource through SMS technology a list of phone numbers and some basic demographic information about the target audience.

Analyze the Target Audience and Media Environment

To start, you need to analyze the target audience and media environment to see if your plan of random texting to get information will work. Somalia is often depicted in the media as a post-apocalyptic warzone where chaos reigns. In reality, it has a robust and thriving telecommunications sector and is home to a number of mobile and Internet providers. The providers have limitations—most are region specific, do not offer roaming services, and face extortion threats from Al-Shabaab. They mostly provide services in major population centers along the coast and largely ignore the smaller inland cities.

Further research reveals that the largest populations of mobile phone users are in the more stable northern regions of self-declared autonomous Somaliland and Puntland, the tip of the Horn of Africa. These regions have large populations and relative stability and safety from Al-Shabaab, and hence are ideal for your initial crowdsourcing project. Your target audience is thus the population of Puntland and Somaliland. Finding the exact number of Somalis who use mobile phones is difficult—some reports suggest about 10 percent of the population uses mobile phones. However, the regional expert (one of the authors) knows that much of the mobile phone–using population resides in Somaliland and Puntland. Additionally, because telecommunications networks are area-specific and do not offer roaming, many mobile phone users use multiple phone numbers or SIM cards so they can access different telecommunications networks. Also, much of the mobile phone population uses dumb phones. Smartphone adoption is on the rise, but far from becoming the norm. Like in other parts of Africa, the target audience uses phones for not only communicating with each other, but for doing monetary transactions and sending each other SMS credits.

Interestingly, information from other area experts reveals that mobile phone numbers in the region follow a set pattern that enables you to derive a list of possible phone numbers. Mobile providers use a specific prefix in a specific area. By knowing a few prefixes in one area from known phone numbers (found on the Internet), you can figure out which prefixes belong to which mobile provider and which area. You can then generate phone numbers by using the prefix and assigning numbers randomly for the rest of the number. Most numbers might be wrong, but a few will work. For example, if a certain town uses the prefix 242, you know all other numbers in the region must be 242-XXXX, where X is a random number. You can then create a list of numbers for that area, such as:

• 242-0001
• 242-4242
• 242-5830
• 242-5840

and so on.

Some more research reveals that radio stations are popular in the area, followed by television. Internet usage is sparse and only about 1 percent of the population seems to have access to it. Finally, another key fact is that the target audience routinely comes into contact with non-governmental organizations and other international organizations. The audience and the region in general are heavily dependent on external assistance for projects and are open to working with foreigners to develop their region as a bulwark against the chaos to their south.

In summary, the target audience and media environment analyses reveal the following key insights:

• The primary mode of social media for the target audience is SMS.
• Mobile phone numbers tend to follow a pattern, so you can generate a list of possible numbers.
• The target audience is willing to work with foreigners, as long as they are convinced that their region will benefit as a result of cooperation.

Keep these insights in mind while designing the crowdsourcing capability.

Design the Platform

The first part of designing the platform involves determining exactly what the platform should look like. The second part involves designing the incentive structure, which, in turn, can affect the look of the platform.

Determine the Platform's Look and Feel

The target audience analysis reveals that the crowdsourcing capability should be an SMS-based platform. The audience regularly uses SMS, and other forms of social media are not widespread. Also, the analysis reveals that you can indeed generate random phone number lists and use existing SMS interaction technology to simply ping the audience for information.

After deciding that you need to build an SMS-based platform, you can start determining how the platform should work. First, you need to generate or acquire a list of phone numbers. Then you need to send an SMS message to the phone number asking the participant for information about the price of a certain commodity. You need to provide an incentive for the participant to respond via SMS. You then assemble the responses and determine which phone numbers elicited a response, and how willing the audience is to participate. Because you are initially generating a list of phone numbers from the pattern, do not expect most of your phone numbers to work. You will likely receive a response from non-working phone numbers saying your message could not be delivered.

The features of the platform are extremely limited by the nature of SMS technology and what you are trying to accomplish. Because you are only interested in collecting simple pieces of intelligence from a wide group of people, you do not need to institute sophisticated social networking or messaging features. The only relevant feature is the ability to send text messages between you and the participants.

The dumb phone in the hands of the participant that receives and sends the SMS is only one component of the platform. The other components include the tool that enables you to send SMS messages to multiple people and the data management system that manages, stores, and displays all the information. In practice, the SMS sending tool and the data management system are usually combined into one piece of technology. Figure 9.2 shows the relationship between the dumb phone, the SMS sending technology, and the data management system.

Figure 9.2 Relationship between SMS platform components

The interface for SMS is the text on a participant's phone and the phone's interface for accessing the SMS. As you will see, the other platform component has its own interface, so you do not need to design any interfaces for this platform.

Determine the Platform's Incentive Structure

You are not asking participants for a lot of information or requiring them to dedicate a lot of their time to your platform. Hence, the bar for incentivizing them is low.

The easiest form of incentive is to offer them a few SMS credits for responding. SMS credits function like a gift card at a store, except instead of cashing in gift cards for goods, recipients cash them in to receive free text messages. Typically, an SMS credit recipient receives instructions through a text message asking him to text a specific shortcode (a truncated phone number that is easy to remember) to receive his credit. Upon texting, he is instantly credited with a few free text messages. Because the target audience uses SMS frequently, free text messages are attractive to them. Because SMS is so ubiquitous, purchasing and delivering SMS credits is easy. Contact (usually through its web page) the prominent mobile phone operator in the region and ask if you can purchase bulk SMS credits. By purchasing in bulk, you receive a discount on each SMS credit. Each SMS credit by itself is very inexpensive, sometimes less than a cent, but they can quickly add up. Plan for about a 20 percent response rate when figuring out how many SMS credits you should buy. For example, assume you plan to text 1000 people and offer each 5 free text messages for responding. Out of the 1000 people, you expect about 200 to respond. If you are giving 5 credits, you need to buy 1000 credits. Make sure that you send each participant a text stating clearly that they will receive 5 free SMS credits upon responding.

Apart from the extrinsic incentive of SMS credits, you can also incentivize the audience through their propensity to work with NGOs to improve their communities. In the text message you send them, communicate that you are asking for price information to help their communities coordinate prices and improve market transparency. If you have the time and resources, contact a local NGO and see if you can get its sponsorship. The audience may already be incentivized to participate with the NGO, and so you may not have to incentivize them through extrinsic items.

Build the Platform

Building the platform only involves utilizing one of many existing SMS sending and management tools. You need to set up the tool, acquire the shortcode and SMS credits, and then start sending and receiving texts. For the SMS sending and management tool, we prefer using the open source and free FrontlineSMS (available at http://www.frontlinesms.com) and RapidSMS (available at http://www.rapidsms.org/). Feel free to try any tool, although we find FrontlineSMS to be the easiest to use. The following steps teach you how to set up FrontlineSMS to “build” your platform:

After testing and purchasing credits, you can start using the tool to deploy your platform and start sending and receiving texts. You can use and customize FrontlineSMS from this point onwards for other platforms. The walkthrough for our platform continues with the following steps:

If you can work with developers or have a developing background, you can also download and manipulate the source code for FrontlineSMS, which is based on Java, JavaScript, and Groovy (a language platform for Java). Modifying the source code enables you to integrate various analytical tools, tweak the program to your needs, change the format for creating and sending messages, and much more. You can do really creative things with FrontlineSMS.

Market the Platform

The marketing campaign for the platform is inherent in the platform itself. You are marketing it by randomly texting potential participants out of the blue and asking them to participate.

However, you may find that participants do not respond, or you want to expand your platform's reach. The target audience analysis reveals that the audience listens heavily to radio. So, advertise on radio stations throughout the life of the platform. In the advertisements, talk up the fact that participants will not only receive SMS credits, but also gain information about prices. Also, partner with a local NGO or an established local organization. The NGO's workers can spread the word or at least reduce the suspicion people may have about getting texts from nowhere. Your text can simply say, “Sponsored by (Name of NGO).”

Manage the Platform

Because the platform does not have social networking or collaborative features, you do not need to hire moderators. You need only one or two people (depending on how many texts you send out) to monitor the FrontlineSMS software and keep track of phone numbers and messages. You also need someone to ensure that you do not run out of SMS credits or credit on your SIM card.

Measure the Platform's Performance

After deploying the platform, start measuring its success rate at finding correct phone numbers and eliciting responses. You can hire developers to integrate analytical tools that measure the performance for you automatically into FrontlineSMS, or you can analyze its performance manually. Example metrics include:

• Number of responses by people
• Number of responses saying “Message could not be sent” because of wrong phone number
• Time and date of responses
• Number of relevant and irrelevant responses
• Number of SMS credits claimed

Wash, Rinse, Repeat

On your initial run of sending texts to participants, you likely will not be very successful. You may find that all the phone numbers you generated are wrong or the pattern finding does not apply to the specific area in which you are interested. You may also find that participants fail to respond, and you may not be sure if they are not responding because the phone number is not valid or because they do not want to respond.

You may need to refine your objective and pick the second option from the list of three you had created earlier. To remind you, the second option is to use the database of phone numbers from marketing companies that do SMS marketing in the region. SMS marketing is typically known as bulk or blast SMS marketing. Blast SMS marketing involves sending out massive amounts, or blasts, of text messages to thousands of phone numbers at a time. Each phone number you blast to receives the same text message. Typically, each blast SMS service has a list of thousands of phone numbers in a certain region. For a fee and a cost per SMS sent, they allow anyone to use their service to send out texts to their phone numbers. Some services even allow the recipients to respond to the advertising SMS. The service forwards the response to you. Some blast SMS services such as Clickatell provide tools to customize their interface for your needs and select from a variety of options. They are easy to use, relatively inexpensive, and have reach in the most remote parts of the world. Another major benefit is that they use phone numbers that are valid and updated so you can be sure that your texts are reaching your audience.

If you are interested in using the service, go to www.clickatell.com and select the International site. Then under Products, select either Easy SMS or SMS Gateway depending on which features you require. You can also use Clickatell through FrontlineSMS. Check out their websites for more information.

Collect Intelligence from the Community for Law Enforcement

Facewatch is a smartphone app and website that British police use to help businesses share information about thieves and other criminals. Facewatch enables businesses to share information about crimes with each other and sort through CCTV pictures of potential suspects to aid identification. Through the app, the police collect all reports and effectively crowdsource the identification of possible criminals from local businesses. If, while sorting through CCTV pictures of possible suspects, a business member recognizes someone as being present when a crime occurred, the business member can then instantly report it to the police through the app. Visit Facewatch's website at http://facewatch .co.uk to learn more.

Facewatch is an ingenious way of crowdsourcing information about criminal activity targeting local businesses and about the identities of possible suspects. However, it does have a few limitations and causes some legal concerns.² One, it is limited to catching petty criminals that affect local businesses and shops and not serial rapists and murderers that terrorize communities. Two, people are poor at identifying possible petty criminal suspects from pictures and they might identify the wrong person. Three, the idea of sharing people's pictures with the suspicion that they are criminals without them knowing it raises concerns about privacy and civil liberties.

This walkthrough explains how to create a smartphone app that, like Facewatch, incentivizes the community to help law enforcement, but with modifications that address Facewatch's limitations. Instead of focusing on petty criminals, this app focuses on catching major criminals such as serial rapists and murderers that affect everyone in the community and not only businesses. The success of the TV show America's Most Wanted has shown that people are more adept at submitting information about a few major criminals, rather than lots of petty criminals. Few people take note of petty criminals and can easily become confused about their identities and whether they have seen them, but are less likely to do so with major criminals. Also, major criminals engender the attention and thus submission of more information by the public, which makes it easier for the police to investigate and connect the dots. A few people will always use the opportunity to submit misinformation or get attention, but the police have always had to deal with that. Additionally, by focusing on only a few people, you are not sharing public pictures of thousands of people. Essentially, this app translates the idea of showing a crude drawing of a major criminal's face on TV and asking people to call in with information to the social media world. The following sections walk through the process of creating such an app.

Define Objective and Scope

Imagine you are working on behalf of local law enforcement in a Western city, such as the Houston metropolitan area, and the objective is to help the local police investigate major crimes including serial rape, homicide, the kidnapping of children, and narcotics trafficking. The impetus is to bolster law enforcement at a time of tight budgets, and to encourage community members to help law enforcement.

Crowdsourcing offers the opportunity to crowdsource information from the public about major crimes and missing children, especially at a fast rate when the crime is still fresh. The police already advertise pleas for information on television and newspapers; however, they would like to increase their reach. Reaching more people will increase the likelihood that they will find people with the right information. If police are not dependent on television shows and newspapers to get the word out, they can ask for information whenever they please. By pushing out information on their schedule, the police can ask for help sooner. People with information will then be able to respond sooner, reducing the chance that incorrect memories will degrade the information over time. Also, young people are increasingly spending more time on the Internet and their smartphones, and less time watching television (and especially the local news shows where police usually ask for tips). A way to reach young people through smartphones would likely increase their participation and willingness to help law enforcement. Additionally, digitizing the information coming in can help police quickly store and manage the information. Lastly, you are concerned about making sure that tipsters can protect their identity while submitting information, and are incentivized to participate.

With these considerations in mind, you can further refine your objective. You need to deploy a crowdsourcing capability in your locality that is attractive to the public, including the youth, that incentivizes users to anonymously submit critical information about major crimes. The capability must then store and deliver the data to police, who can use it to further their investigations. With a clear objective in mind, you can continue on and analyze the target audience. Note that information from the target audience analysis could cause you to revisit and modify the objective.

Analyze the Target Audience and Media Environment

The target audience is the people that live in the Houston metropolitan area, and especially the youth. However, that includes nearly 6 million people. Deploying a capability for the first time to such a large number of people is difficult—your message may become diluted, or so many people may participate that your capability gets overwhelmed. So, consider focusing your efforts to perhaps only the city of Houston. The city has 2 million people, which is a large but more manageable number of people.

After choosing the target audience's location, you can analyze their demographics. Some quick research on the Internet reveals that Houston's population is fairly young and diverse, with a significant part being Hispanic. Thus, to reach the widest audience, whatever capability you build should be available in both English and Spanish and appeal to young people.

Analyzing the target audience also involves analyzing their media environment and how they use media. Further research and insight from an area expert (one of us grew up there) reveal that Houston's population is similar to the population of other major American cities. Much of the population has access to the Internet, and smartphone technologies are popular, especially among the youth. However, some parts of Houston are mired in poverty and have less regular access to Internet technologies. Of those who do use the Internet, many use social media including Facebook and Twitter. Analysis of social media use in the United States suggests that 18- to 40-year-olds use social media the most. Also, a significant part of the population and likely much of the youth regularly use smartphone technologies, including the iPhone and Android phones.

Due to the widespread use of Internet and social media, Houston's media environment is saturated. People there consume media in numerous ways and are constantly being bombarded with commercials about new ways to spend their time and resources. In a given day, people living in Houston can watch cable television, access Facebook on their iPhones, and read numerous newspapers. Getting their attention will be difficult in such an oversaturated media environment. Also, they will have to be incentivized in some way to replace one form of media use with your capability.

In summary, the target audience and media environment analyses reveal the following key insights:

• Much of the target audience is young and Internet-savvy.
• A significant part of the target audience regularly speaks Spanish.
• Smartphones are very popular among the target audience.
• Getting the target audience's attention will not be easy.

Keep these insights in mind while designing the crowdsourcing capability.

Design the Platform

The first part of designing the platform involves determining exactly what the platform should look like. The second part involves designing the incentive structure, which, in turn, can affect the look of the platform.

Determine the Platform's Look and Feel

The target audience analysis reveals that the crowdsourcing capability should be a smartphone application like the apps available in the Apple and Android app stores. Much of the target audience regularly uses smartphones. Also, because smartphone applications are very popular, the audience will likely respond approvingly to a smartphone app as opposed to simply another website.

After deciding that you need to build a smartphone app, you can start determining what the app should do and, specifically, which features it should have. The app should allow law enforcement to push out information about a crime, such as the photo of a suspect, the last known whereabouts of a missing child, and other details to participants who have downloaded and installed the app. The participants should then be able to consume the pushed information and respond to it with tips. In response, the participants should receive some type of extrinsic and/or intrinsic incentive, especially if their tip proves useful. Participants should also be able to share the app and the police's requests with others. Figure 9.3 represents an overview of the app.

Figure 9.3 Law enforcement app overview

The features will be heavily dependent on how you would like the participants to use the app in a way that best helps your objective but also keeps participants interested. The following list describes some of the features you can implement and the reasoning behind them:

• Messaging between police and participants—The most obvious feature, the ability to send messages between the participants and police, is pivotal. Police can push a message to participants about the crime or the suspect. The participants can then respond with whatever tips they may have. The messages can include all sorts of multimedia, including pictures, text, and video. The police can also message only a single participant. For example, if a participant submits useful information that leads to a breakthrough in the investigation, the police can send that individual a thank-you message for his or her efforts.
• Notifications—Notifications are the messages you receive when you are not using your smartphone that remind you that you have mail, or need to make a play in Words with Friends, or have a new text message. Participants who download the smartphone app have the option of turning on notifications for the app. With notifications on, when police push information about the crime, participants will receive a small reminder alerting them that police are requesting new information.
• Location—Many smartphone applications ask the users for permission to track their physical location. If the platform asks participants for their location, which they can always decline to do, you can track where they are located. By analyzing the location of participants you can determine whether your platform is especially popular in some parts of the city, or completely absent in other parts. Also, you can use a participant's location to validate the tips. Say that a series of related crimes happen in the south part of the city. A participant then submits information about the crimes, suggesting that he or she is near the crime while the information is being submitted. However, the location may reveal that the participant is actually in the north part of the city. You can then figure out that the participant is likely lying or, at best, not telling the complete truth.
• User accounts—After installing the app, each participant has to create a user account. Participants must specify a username, which the app will recommend is not the participant's real name. Participants will then also have the option of submitting information about their age, gender, and race to understand what parts of the audience are using the app. Participants will also have the option of creating a password for the app.
• Social sharing—To ensure that the largest number of people possible hear about the police's requests, the app will enable participants to share information with people in their contact lists. The participants will have the option of sharing the requested information with anyone else through a text message or an e-mail. The shared message will also contain instructions for the recipient to either download the app, or call the police tip-line. For example, a participant may see the picture of someone and think he recognizes it, but is unsure. He may feel that a family member might have a better idea, and so he can immediately message the picture and instructions to that family member. The family member can then check out the picture and respond to either the participant or directly to the police.

The app that participants use is only one component of the platform. The other components include the data management system that handles all the data and the website that enables the police to access the crowdsourced intelligence. The data management system should generally allow for the delivery and storage of all messages and metadata about the messages. The data management system should also ensure that it routinely deletes information about participants and does not store any information that personally identifies them, according to legal and privacy obligations that the police face. Figure 9.4 shows the relationship between the app, the data management system, and the website.

Figure 9.4 Relationship between app platform components

The website that police access on their computers should have at least the following features:

• Messages—All the communication including the text and media shared between the police and the participants must appear on the website. The messages can be organized by case. Police users can click on a case and then see all the messages and information pertaining to the case, including the usernames of the participants who submitted the messages. Police users should then be able to click on the username and submit a message directly and only to that participant.
• Updates—The police user should be able to post new case information on the site and then push it through the message feature onto the smartphone apps.
• Sharing—Police users can create accounts for themselves and share the interface with other members of the police force, including analysts.

After outlining which features the app and the police website should have, you could start designing the graphical user interface (GUI) of the app and the website. However, we recommend thinking about the incentive structure first, because that can affect the features available on the app and, thus, how the app should display the features.

Determine the Platform's Incentive Structure

Looking at prior ways of incentivizing the general population to submit information to police regarding crimes is a good starting point for determining your incentive structure. Usually, law enforcement, community organizations, and families of victims offer cash rewards for information. The cash rewards depend on the seriousness of the crime and the need for new information.

Cash rewards are important extrinsic incentives that routinely encourage participation and are easy to implement for the app. Small cash rewards in the range of a few hundred dollars are likely adequate for incentivizing people. In the case of very serious crimes or missing children where time is of the essence, the police can increase the amount of the cash reward by a few hundred dollars. Delivering the cash reward is also a simple endeavor. The website on which the police access the incoming messages stores each message and the name of the person who submits it. If a message proves useful, the police can then use the app to communicate with the participant who submitted it. Through the app, the police can tell the participant how to collect the reward—most likely by calling a phone number and providing his name and address, or coming to the police station with his phone to prove that he was the one with the username.

Intrinsic incentives also come into play—specifically, the pleasure people gain when they feel they are contributing to their community and helping their neighbors. The app enables people to be useful, and the police can routinely send thank-you messages to participants through the app and let them know they were helpful. Even in cases where the information was not especially helpful, the police can still send messages to participants now and then thanking them for trying to help. The message can say something like, “Thank you for the information. We appreciate your service to your community.” In the case of crimes or events where the likelihood of repercussions to the informant is nonexistent, the police can even publicly thank the informant. Public thanks and kudos are a strong intrinsic incentive.

Determine the Platform's Look and Feel, Redux

The incentive structure does not impact the design and structure of the app or the website. Thus, you do not have to reassess ideas for the GUI and can continue as planned.

You can design the GUI in countless ways. Whichever way you choose, it has to ensure that participants and police can access the features of the platform in an intuitive way. Figure 9.5 shows one example of how the app's GUI could look, or in other words, what the app could look like to a participant. It shows the different windows, screens, or interfaces that participants can access in the app.

Figure 9.6 shows one example of how the GUI of the website that police use to access the information could look.

Figure 9.5 Example GUI for app

Figure 9.6 Example GUI for police website

Build the Platform

Building the components of the platform is a simple technology problem that requires contracting out work to developers. You can also use services such as AppMakr (www.appmakr.com) and Magmito (www.magmito.com) that help you build and deploy smartphone apps. We prefer going to developers rather than doing it ourselves because they can ensure that the software behind the platform is secure and that we get all the desired features without cutting corners. Ideally, you should work with a team of developers that work together on all of the components of the platform, rather than farming out different components to different developers. With the latter strategy, piecing the platform components together can become difficult.

Providing the developers with example GUI graphics and a list of the features you need significantly increases the chance that the end product fits your desires.

Market the Platform

Because the target audience's media environment is saturated, getting the audience's attention and getting them interested in the app will require a robust marketing effort. The effort should consist of an initial marketing push and a sustained marketing effort throughout the app's existence.

The initial marketing push should introduce the app to people and encourage them to sign up and try it out. The most important step for the initial marketing push is naming the app. You do not want to pick something too cheesy, nor something that obscures what the app does. For lack of a better idea, we will go with “CrimeSpotter.” After naming the app, list it in the Apple and Google app stores. Listing it in the app stores involves going through an approval process, which may take some time and require modifications to the app.

Most of the target audience regularly consumes traditional media. So, talk about the app on television news shows, and place advertisements on television, radio, and in magazines and newspapers. Contact Apple and Google and encourage them to help you market the app, because doing so would improve their public perception. Also, because much of the target audience regularly uses the Internet, advertise on the Internet. Launch a website that explains, in simple terms, what the app does and why people should use it. Run ads on Facebook targeted toward young people living in Houston. Contact celebrities and encourage them to tout the ad on television and at concerts.

The sustained marketing effort includes regularly sending messages to participants through the app that encourage them to tell their friends about it. You do not want to send messages too regularly, otherwise you will annoy them—once a month may be enough. Also, analyze who is signing up for the app based on the demographic information you are collecting through the app and use it to tweak the marketing effort. If you find that only females are signing up, start advertising in magazines for men such as Men's Health. If the participation rate follows the technology adoption curve and your participation rate is high, you do not need to advertise as much.

Manage the Platform

Because the app does not have social networking or collaborative features, moderating the platform will not require many resources. Dedicate a few analysts or police officers to monitor the police website to ensure people are submitting relevant information. If you find that you are receiving only spam and nonsense, you may want to shut down the app. Also, have the dedicated analysts or officers message out reminders and thanks to participants.

Measure the Platform's Performance

After launching the app, start collecting metrics for its performance. Of course, you will want to think about what metrics to use before you even commission its design and building. The data management system component of the platform should house some performance tools that measure statistics about the app's usage. You can also acquire data from Apple and Google about the app. Example metrics include:

• Download rate of the app over time, compared with the schedule of marketing pushes
• Number of incoming messages, compared with other ways of getting tips
• Number of successful tips
• Number of cash rewards given out
• Time and day people submit information, to assess when participants are most active and when notifications should be sent out
• Demographic data about who is using the app
• Stories in the media about the app and participants' assessments of it

Wash, Rinse, Repeat

Undoubtedly, your platform will fall short in some ways. Participants may demand certain features, or you may need to tweak the marketing efforts, or even further refine your target audience. Use the performance data to identify areas that you need to improve. You may even consider hosting a focus group a few months into the app's existence to talk with participants about what they do and do not like about the app. Use the results of the focus group as a guideline, not commands. Sometimes people do not know what they are talking about and may suggest very stupid things.

While writing this, we came up with a few ways of improving the app through its subsequent iterations and updates:

• One major flaw jumped out immediately—with the existing app, participants cannot look at past messages. You never know when they might remember something or stumble onto the information you need. You will need to integrate an option into the app that enables participants to browse past messages.
• Create a companion website for the target audience that enables people without a smartphone to participate.
• Coordinate the sending of messages with television and radio advertisements to ensure that participants know a new message is up.
• Drastically increase a reward every now and then to keep participants interested and excited over the long term.
• Enable participants to upload videos through the app, and even turn their smartphone into a live video camera that the police can view in case a crime is occurring near the participant.

Collect Intelligence to Bolster Disaster Relief

Natural disasters and humanitarian crises are events that require crowdsourcing information from affected victims. Without on-the-ground information about the crisis and its effects, relief workers and community members are helpless. Responders and relief organizations use the crowdsourced information to identify areas of distress and coordinate supplies. News organizations and community members use the information to share warnings about road closures and locations of shelters, and raise awareness about the crisis.

However, not all information is useful. The best and most useful information is the type that is timely and valid. During or immediately after a crisis, responders need information as soon as possible so they can coordinate their efforts and reach people in need quickly. For example, responders who find out days after a tornado that a baby is missing will not be able to help the baby, who will likely have died since. Crowdsourcing relevant information through social media offers the ability to significantly improve the timeliness of the information. Also, information is only good if it is true and verifiable. People take advantage of crises to commit crimes against victims and responders, and wreak havoc.³ They may send information asking for help, but instead set traps to kidnap relief workers. The ability to validate incoming information can go a long way toward ensuring that relief efforts are not diverted toward false and possibly dangerous ends. Social media technologies offer the ability to cross-validate incoming information and quickly employ others to validate questionable information.

Crowdmaps are websites where people can submit information that is then displayed on an interactive map, similar to how Google Maps displays information about stores and tourist attractions on its maps. Crowdmaps are ideal for quickly capturing a lot of reports and pictures during times of crises and displaying that information on a map. Numerous individuals and groups have used crowdmaps to track disasters and coordinate disaster relief. Some have helped first responders coordinate disaster relief and identify victims. However, the vast majority are either rarely used, ignored, or not shared with the proper authorities because of numerous reasons. This walkthrough describes how to successfully implement a crowdmap-based crowdsourcing intelligence platform to collect intelligence during crises to bolster disaster relief and document problems. The following sections walk through the process of designing, creating, deploying, and maintaining a crowdmap.

Define Objective and Scope

Imagine you are charged with helping coordinate disaster relief in Haiti, a few months before the hurricane season is expected to start there. Your main task is to collect information about the hurricane and its effects from affected Haitians and provide the information to disaster responders so they can better plan their actions. A crowdmap is the ideal crowdsourcing platform for collecting and managing intelligence during crises.

Considering past crowdmap deployments for the purpose of disaster relief in Haiti and other places can help further refine the objective and its scope. Although some crowdmaps prove enormously useful to the victims and responders, most crowdmaps fail because no one knows about them. If people do not know that a crowdmap exists, they will not submit reports to it. Responders also need to know about the crowdmaps. Crowdmaps that are popular with the affected population will receive and host a lot of information. However, if responders do not use the information, the crowdmaps become a simple exercise in documenting damage and problems. Therefore, your crowdmap deployment must feature a significant marketing effort. Additionally, popularizing a crowdmap and populating it with lots of information is not enough. False and nonsensical information can significantly harm, or at the very least prove useless to, responders. A successful crowdmap deployment must manage and make sense of the crowdsourced information, especially if responders will use it to plan their actions.

With these considerations in mind, you can further refine and restate your objective. You need to deploy a well-known crowdsourcing platform in Haiti during disasters that collects reports from the affected population about problems they face and the help they need, manages and prioritizes the information, and then pushes it to responders quickly for their use.

Analyze the Target Audience and Media Environment

The region you are interested in largely determines the target audience for the platform. Your target audience is obviously the population of Haiti, which has about 10 million people. The number of people is far too large to deal with for an initial crowdsourcing deployment, so you will have to further refine the target audience.

The easiest option is to pick a major city and focus on the population there. Populations in major cities tend to have the most advanced technologies and are more likely to use social media technologies and participate in crowdsourcing. Haiti's capital city, Port-au-Prince, seems ideal because it is one of the most advanced cities in Haiti. It has suffered significant damage in past crises and so will likely suffer the consequences of future earthquakes and hurricanes. Also, it has almost 1 million people. The number of people is still quite large, but much fewer than 1 million will likely even have the ability to participate in the platform. Keep in mind that a significant portion of Port-au-Prince's population does not have access to social media technologies. Indeed, research shows only 35 percent of Haiti's population uses dumb phones. Compared to other countries, the number is small, but large enough for the platform's purpose. Therefore, you can select the population of Port-au-Prince as the target audience for the platform.

The target audience's media environment is heavily dependent on traditional media technologies. Haiti has numerous television and radio stations. Overall, radio remains the most popular source of media for Haitians. Only about 10 percent of Haitians have access to the Internet. However, access to mobile technology is rapidly growing. Haiti is also deploying 3G networks, which will further spur use of social media technologies. Also, the Haitian diaspora has a significant influence on Haiti's way of life, and thus the media environment. The diaspora has even become involved in past crowdmap deployments in Haiti. It is unclear from the analysis how many Haitians know about crowdmaps, but we expect they will participate if they find out about them. Disaster response in Haiti suffers from significant problems, resulting in the loss of hundreds of thousands of Haitian lives. Thus, it is safe to assume that the target audience will welcome ways of improving disaster response.

In summary, the target audience and media environment analyses reveal the following key insights:

• A significant portion of the target audience regularly uses dumb phones and is likely familiar with SMS services.
• Traditional media dominates Haiti's media environment, and radio is the best way of reaching people.
• The target audience is already incentivized to try out tools that will improve disaster response.

Keep these insights in mind while designing the crowdsourcing capability.

Design the Platform

The first part of designing the platform involves determining exactly what the platform should look like. The second part involves designing the incentive structure, which in this case will likely not affect the look of the platform due to the nature of the crowdsourcing technologies in question.

Determine the Platform's Look and Feel

The target audience analysis reveals that the crowdsourcing capability should be based on SMS technology. The platform should allow the target audience to submit SMS messages to a centralized data management system detailing the help they need and where they are. The system should then take the information, and through either automated tools or human labor, clean up, verify, translate, and prioritize the pleas for help. Responders should then be able to look at the information and use it to plan their actions, and even update the crowdmap or contact the person who submitted the information. Figure 9.7 represents an overview of the platform.

Figure 9.7 Disaster relief platform overview

The platform has two components. The first is the SMS capability that allows the target audience to submit information to the platform. The second is the crowdmap, which stores, manages, and displays the crowdsourced information. The platform is relatively simple and has the following few features:

• Messaging—Obviously, participants should be able to send a text message to the platform with details about where they are and what help they need.
• Location—The amount of location information you receive depends heavily on the settings of the participants' cell phones, the policies of the local telecommunications providers, and other technological items. Typically, the location is estimated by analyzing which cell towers the message bounced off of on its way to the crowdmap. The location you receive may not be exact, but it will good enough. Remind the participants to specify where they are in the body of the text message.
• Social sharing—To improve relief coordination, the collected messages must be shared with the responders. The platform will enable all responders or coordinators to immediately see what type of help people need, where they are located, and when they need the help. The coordinators should also be able to quickly share the information with others organizations or persons.

A simple set of features is a clue that the user interface for the platform will also be relatively simple. The user interface for SMS is only text on a person's phone. You do not need to design any graphics. Numerous user interface options exist for the crowdmap. You can design one from scratch, adapt existing ones, or use a template with minimal changes. Assuming available resources and time are low, you should opt for doing the least amount of work possible. Fortunately, Ushahidi provides easy-to-use crowdmap templates with the desired features for free. The template determines what the interface will look like. Thus, unlike other platforms, you have to do almost no work in regards to designing the look and feel of the platform. Obviously, more custom and expensive options offer more features, but for our purposes, time, and budget, Ushahidi's crowdmaps will do.

Determine the Platform's Incentive Structure

Haitians have suffered massively due to disasters, which is an unfortunate fact that significantly simplifies your platform's incentive structure. One of the strongest intrinsic incentives people have is the need to feel secure and to be free from harm. The need for security is so strong that you do not need to consider other incentives. If deployed and used correctly, the disaster relief platform can considerably reduce Haitians' insecurity and likelihood of harm during disasters.

However, it takes effort to convince the target audience that the platform will indeed help them. If the target audience believes the platform will not help them, they will not use it. A convincing marketing campaign is essential for articulating the incentive for the audience. Also, an incentive is only as good as its delivery. Even if you convince the audience that your platform will help them, if you then fail to help them through the platform, you will not be able to deploy another platform there again. Even during a disaster, word will spread among the affected communities about the failures and accomplishments of the platform. Make sure the platform works, otherwise the effect of the intrinsic incentive will disappear and the participation rate will plummet.

Build the Platform

Building the platform involves creating and deploying a crowdmap, and syncing it with an SMS management and blasting tool such as FrontlineSMS or Clickatell. The process is fairly easy and does not require much technological skill. The following steps describe the process:

You will now need to set up the SMS component of the platform. You can use FrontlineSMS to do so, but it is tricky to use for this case so we prefer using Clickatell:

Figure 9.8 Screenshot of crowdmap dashboard settings

Market the Platform

Crowdmaps succeed only if people know about them, and so you need to put a lot of resources into marketing the platform. Ideally, you want to routinely advertise the platform over the course of many months and even years till it becomes common knowledge among Haitians. The target audience's media environment is heavily dependent on traditional media. Advertise the platform on radio and on television. Make a jingle out of the shortcode so it becomes memorable. Consider that almost everyone in the United States knows to dial 911 in the case of emergencies. Similarly, everyone in Port-au-Prince should know to text to shortcode XXXXX during a disaster when they need help. The name of the platform might as well be the shortcode because it is the shortcode that is the most important piece of the platform. If the audience does not know it, they and the responders will not be able to use the platform.

To ensure such ubiquity of your message, you will likely need to partner with Haitian organizations. Contact the Haitian government, the Red Cross, and other NGOs and convince them to market your platform. They can help tell key influencers in communities about your platform and print and distribute flyers. Also, consider contacting Digicell, the largest provider of mobile services in Haiti. Digicell can help market the platform and shortcode, which in turn will make its product, the mobile phone, look more useful and attractive to its potential customers.

You can also use blast SMS to market the platform. While building the crowdmap, you had to sign up for Clickatell, which you can also use to send text messages to Haitians. Every few weeks, blast SMS reminders about the platform to the target audience to slowly build up the number of people who know about it.

Also, consider running drills to both market and test the platform. Work with NGOs and local government offices in Port-au-Prince to stage moderate-sized drills where you simulate a disaster and ask a small portion of the target audience to role-play as disaster victims and use the platform. The drills will help the audience appreciate what the platform does, and it will generate buzz and raise the audience's interest.

Manage the Platform

Moderators will play a significant role in managing the crowdmap and coordinating with responders. They will categorize, validate, translate, and push the information. You will likely need to hire at least six moderators to manage the platform during a crisis. The moderators should know English, French, and/or Haitian Creole, which are the most commonly used languages among the target audience. The moderators should work in shifts, with at least two working together at the same time. The following list describes the most important of the moderators' duties:

• Categorize the incoming text messages into the categories you delineated when setting up the crowdmap.
• Validate or approve messages by cross-referencing them and seeing if they make sense according to what is generally happening on the ground.
• Send alerts to responders about messages that ask for immediate assistance.
• Translate incoming messages by themselves or through the use of other tools.
• Manage the crowdmap and SMS credits settings to ensure everything is working fine and they have enough credits to receive texts.
• Update the crowdmap with information they pick up from other sources, and if certain issues are resolved by responders.

If you cannot find moderators with all of the appropriate language skills, or if they are being overwhelmed with messages that need to be translated, then you can use crowd labor to translate. Chapter 10 describes how to use crowd labor to solve problems such as the translation of language containing slang, abbreviations, and idioms.

Fortunately, Ushahidi's crowdmap software already features all the moderating tools you need. Make sure to train the moderators on the crowdmaps. Drills are ideal for training and testing the moderators. The moderators will spend the majority of their time in the Reports section of the crowdmap dashboard. See Figure 9.9 for a screenshot of the Reports section. The Reports section provides moderators with the tools to track incoming reports, and validate and approve them. Moderators can also click on each individual report and edit them to add more information. Moderators can also go to the Messages section where they can classify messages as reports and even send messages.

Figure 9.9 Screenshot of crowdmap Reports section

Measure the Platform's Performance

After deploying the crowdmap, collect the metrics to measure the platform's performance. You likely noticed as you set up the crowdmap that it features several tools to collect and make sense of the data. Use these and other tools, and moderators to assess the platform's performance. Example metrics include:

• Number of texts received
• Number of different phone numbers texting in
• Number of victims identified compared to past disasters
• Number of victims helped compared to past disasters
• Increase in response time compared to past disasters

Wash, Rinse, Repeat

The success of the crowdmap depends heavily on the marketing push and how much of the target audience knows that the crowdmap actually exists. Do not be surprised if only a few participants text in during your first crowdmap deployment. You may need to sustain your marketing campaign for a long period of time before the target audience realizes that they need to text in to a specific shortcode to get help. Getting buy-in from the local government may be critical.

Also, we addressed only a few of the features that crowdmaps offer. So far, we have restricted how Haitians can send messages to the platform to only SMS. For future deployments, you may need to expand the modes of communication to include e-mail and Twitter. Increasing the ways participants can communicate with you leads to an increase in the intelligence you receive. To compensate for the increase, you will need to hire more moderators or employ crowdsourcing to help with some of the problems you may face.

Chapter 10 focuses on building crowdsourcing platforms to solve problems, including the translation of hard-to-understand languages and much more.

Summary

• Crowdsourced intelligence is the information you collect from crowdsourcing platforms known as intelligence collection platforms.
• The type of intelligence you receive is dependent on the form of social media participants use to submit the intelligence. For example, if collecting intelligence through SMS, the intelligence will be in text format.
• Direct intelligence tells you about a place, an object, or a person, and other than validation and sorting, does not require additional analysis.
• Indirect intelligence tells you about the things surrounding a place, an object, or a person, and requires analyses and correlations with other intelligence to derive insights about the place, object, or person.
• Intelligence collection platforms suffer from a few limitations:
  • The target audience may be uncooperative because they do not want to help or they are afraid to participate.
  • Participants may face certain risks due to the somewhat open nature of social media communication.
  • Participants may send you incorrect information, either intentionally or by accident.
  • You may misinterpret data and draw wrong conclusions from it.
  • The platform may change the behavior of the target audience, which may impact what kind of intelligence you can get.
• Compared to other types of crowdsourcing platforms, intelligence collection platforms are typically run for shorter periods of time, require less elaborate incentive structures and social networking features, and are primarily deployed in hard-to-reach areas where SMS use is high.
• You can use intelligence collection platforms to collect information about the target audience from hard-to-reach areas through SMS networks.
• You can use intelligence collection platforms to collect information about crimes from the community through a smartphone application.
• You can use intelligence collection platforms during disasters to collect information from affected victims through a crowdmap and SMS network.

Notes

1. Burnett, J. (2011) “Mexican Drug Cartels Now Menace Social Media.” NPR. Accessed: 18 August 2012. http://www.npr.org/2011/09/23/140745739/mexican-drug-cartels-now-menace-social-media?sc=tw&cc=share

2. Badger, E. (2012) “How the Cutting Edge in Crowdsourced Crime Fighting Could Do More Harm Than Good.” The Atlantic Cities. Accessed: 18 August 2012. http://www.theatlanticcities.com/technology/2012/07/ how-cutting-edge-crowdsourced-crime-fighting-could-do-more-harm-good/2626/

3. Jordan, M. & Gauthier-Villars, D. (2010) “Kidnapped Aid Workers Released in Haiti.” Wall Street Journal. Accessed: 18 August 2012. http://online .wsj.com/article/SB10001424052748704349304575116011173049380.html