On Friday, January 22, 2016, the Ontario Court of Justice released reasons in R. v. Elliott, 2016 ONCJ 35. The case involved allegations that the accused engaged in criminal harassment of two women by repeatedly communicating with them over Twitter. The case is interesting in a number of respects, not least of which was the court’s struggle with how to address evidentiary challenges with social media.

A Little Background

The case stemmed from a dispute that took place over Twitter between the accused and, among others, the two complainants. The accused met one of the complainants once, but never met the other. An analysis of all of the tweets that were involved goes beyond the scope of this post. Suffice it to say that the exchanges were not always a model of civil discourse – on anyone’s part.

In order prove criminal harassment under section 264 of the Criminal Code, the prosecution was required to establish:

  • the accused repeatedly communicated with, either directly or indirectly, with the two complainants
  • the complainants were harassed
  • the accused knew the complainants were harassed or was reckless or wilfully blind as to whether they were harassed
  • the complainants feared for their safety (physical or psychological)
  • the fear was reasonable in all the circumstances

The court had no trouble concluding that most of the elements of criminal harassment were established with respect to the complainant with whom the accused had had personal contact and who the accused engaged with the most (directly and indirectly) over Twitter. However, the prosecutor was unable to prove that the harassment, however genuinely felt by the complainant, was reasonable.

The prosecution relied on the number of tweets and that the accused continued to participate in a public online discussion of various topics in which the two complainants were involved even when they asked him to stop. But the court concluded that there were no tweets that were of a violent, sexual, or irrational nature. In the absence of that type of conduct, the court appeared to think that the complainant’s participation in shaming the accused should be part of the context in which the volume of tweets should be understood. The court held: “The main premise that I find unreasonable is [the complainant’s] perception that she could tweet about topics but not be exposed to [the accused’s] tweets (however spurious and invalid) about the same topic – even if the topic was him.”

Listening in on the Public Square

The allegedly harassing tweets and the context in which the tweets occurred were, of course, central to the entire case. However, no single tweet or group of tweets were alleged to constitute the harassment. Instead, the harassment was alleged to occur because of the total volume of tweets. Since one of the complainants had blocked the accused on Twitter, some of the tweets that allegedly constituted part of the harassment involved those that were sent to others, such as the complainants’ followers, or tweets that contained hashtags that the complainants used.

One of the complainants likened Twitter to a public square. With all of the shouting in the public square, how would the prosecution select the tweets and demonstrate that they were harassing in context? Initially, the police simply searched the Twitter platform for tweets but the number of tweets were temporally limited and this type of search would not capture erased tweets. So the police resorted to employing social media listening software, which is used by organizations to track trends and to understand how potential customers are engaging with brands and ideas. Using this social media listening software, the investigating police officer:

  • looked for the conversation between the accused and the complainants by searching for tweets in which they used each other’s Twitter handle
  • looked for tweets that contained certain hashtags that the accused or the complainants either created or followed

Importantly, because the tweets were available to anyone with a licence to use the social media listening software, the court accepted that it was unnecessary for the police to obtain a warrant to gather that evidence. Although not discussed by the court, it does not appear that a user of Twitter could make a viable argument for a reasonable expectation of privacy based on the current version of the Twitter privacy policy. The privacy policy specifically notifies users that the Twitters services “broadly and instantly disseminate your public information to a wide range of users, customers, and services” and that this information may be delivered and used by third parties to analyze the information for trends and insights.

Proof the Accused Sent the Tweets

The charges of criminal harassment did not require the prosecution to prove that the accused meant what he said in the tweets. But it was necessary to prove that the accused sent them. But the fact that they were sent by the accused was hearsay. At a minimum, it was necessary to establish a link between the accused and the handle from which the tweet allegedly originated and then to establish that the tweets were authentic in the sense of not having been altered.

The court took a common sense, although perhaps not particularly rigorous, approach to this issue. The court noted that some of the tweets attributed to the accused related to a dinner that the accused had with one of the complainants. This provided circumstantial evidence that the accused was tweeting from that handle. Since there was no evidence anyone else had access to the accused’s handle or had access to his account, and the accused did not allege the tweets had been altered, the court found it was not necessary to prove that the accused sent each tweet.

The Fragility and Completeness of Electronic Evidence

There were numerous problems with printing out the tweets obtained through social media listening software. The searches using the social media listening software did not always yield complete records of the tweets. The tweets were garbled with punctuation being displayed as symbols.  Links and attachments were not available.

To remedy the deficiencies, the prosecution created electronic files that showed the tweet as it appears in Twitter. However, using this method meant that the court was being connected to Twitter through the Internet. The court encountered striking examples of how access to the evidence could be impeded in real time during the trial. For example, at one point, the prosecution could not open a tweet because the complainant, who was testifying, had locked her account and made it private the day before she testified. The problem re-occurred when the defence counsel attempted to open the links only to discover the complainant had again blocked access to her account. A similar issue happened to the trial judge when hew as attempting to review the evidence while preparing his reasons for judgment.

Ultimately the electronic evidence had to be printed in order to create a stable record of the evidence that was introduced at trial.

Selectivity in Relying on Search Results

Another problem for the prosecution was establishing the context for the tweets. If only the tweets produced by the search results had been authenticated and introduced into evidence, the prosecution would have failed to prove the context of a tweet if the court could not give meaning to the content of the tweet without resort to other tweets or the links in the tweets.

The court took a common sense, although again not a particularly rigorous approach, to the evidence. The court held that if there were other tweets surrounding the tweets sent and received by the accused that provided context and were printed and available in the exhibits, he could rely upon those to give the tweets the appropriate context.

Twitter Discourse is Not a Free-for-All

Although the charges against the accused were dismissed, the court did not find that anything goes on Twitter. Clearly, messages that are threatening can be objectively harassing. It is also possible that the total volume of messages could be objectively harassing if the complainant tells the user that the volume is harassing, blocks the user and does not engage with the user further, and the user persists.

In future cases, it may be relevant for the court to consider whether the complainant sought recourse through Twitter. Twitter recently updated its Rules to address abusive behaviour. In particular, the Rules prohibit behaviour “that harasses, intimidates or uses fear to silence another user’s voice.” Twitter will consider the following factors when considering whether there is harassment: (i) whether the primary purpose of the account is to harass or send abusive messages; (ii) whether the reported behaviour is one-sided or involves threats; (iii) whether the account is being used to incite others to harass another account; and (iv) whether the account is sending harassing messages from multiple accounts. Twitter has a number of tools that it makes available to report violations.